DTI Custom Publication, 2009

From a Leading Provider inEmbedded Computing and Switch Productsfor the Government and Military Markets

Diversified Technology, Inc.

> COTS GotchasWhat to Look for When Selecting a COTS Provider

> 10G Ethernet - AdvancedTCA SwitchingAn AdvancedTCA Switching and Protocols Primer

> COM-onomicsThe Economics and Use of COM Express in Embedded Applications

> Mobile Power for Mobile ForcesOn-Board Vehicle Power

> High Availability Load BalancingWith HAProxy on AdvancedTCA Hardware

1.800.443.2667 • [email protected] • http://www.dtims.com

CPU Boards - Switch Blades - Rackmount Systems - Modular Platforms

Computer Modules - System Integration - Custom Designs - Outsourcing Capabilities

Three Decades of Embedded Solutions

All trademarks and tradenames are the property of their respective owners.

High PerformingPlatform Systems

Diversified Technology, Inc. (DTI) designs and manufactures embedded

platforms, communication servers, and subsystems for the mission critical and

non-critical demands of the government computing market. DTI Solutions,

delivered on a variety of form factors, offer long life support for multiple

government agencies involved with protecting global and domestic interests.

Embedded computing in the military will continue to drive DTI to new COTS and

proprietary solutions to support the government's defense initiatives.

> Find out more at www.dtims.com

• Communication Blade Servers based on OpenArchitecture Standards such as CompactPCI®, VME and AdvancedTCA®, and featuring thelatest processing and switched fabric technologies.

• PCI/ISA CPU Boards and Rackmount Systemsutilizing high-slot count backplanes forthird-party cards in the end solution.

• Embedded Computer Modules based on ETXand COM Express provide computing powerin reduced physical dimensions.

• Outsourcing and Customization Options Available through DTI’s In-House Engineering.

Embedded Computing Solutions

AdvancedTCA • VME • CompactPCI • PCI • ETX • COM Express • Custom Design and Manufacturing

Page 3DTI Wired : Get Connected with Diversified Technology, Inc.

Inside This Issue

COTS Gotchas..........................................................05Key Areas to Consider when Selecting A COTS Provider

10G Ethernet ............................................................15An AdvancedTCA Switching and Protocols Primer

COM-onomics ..........................................................22The Economics and Use of COM Express in Embedded Applications

Mobile Power for Mobile Forces ............................26On-Board Vehicle Power (OBVP)

The Cohesive ApproachHow to Achieve High Availability............................32Load Balancing with HAProxy on AdvancedTCA Hardware

Message from the President ..................................04

VME Solutions..........................................................12

Custom Designs ......................................................13

Employee Spotlight ................................................13

About Diversified Technology, Inc. ........................45

Joe Speak ................................................................45

DTI Tradeshow Schedule ........................................46


476 Highland Colony Parkway

Ridgeland, MS 39157

(800) 443-2667

[email protected]

www.DiversifiedTechnology.com

Pat BusbyPresident

Barrie McArthurVice Chairman - Special Projects

Steve CravenVice President, Program Management

Donald GermanyVice President, Systems Engineering

Ken MartinVice President, Finance

Joe McDevittChief Technology Officer

Gary SmithVice President, Operations

Keith VarnerVice President, Solutions Group

Patrick WelzienVice President, Engineering

Tom WittVice President, Business Development

About Diversified Technology, Inc.

Diversified Technology, Inc. (DTI), an ErgonCompany, was founded in 1971 and focuseson primary markets of Communications,Government / Military, Commercial, PowerElectronics, On-Board Vehicle Power andSimultaneous Field Radiation Technology. Asan embedded hardware, software andsystems company, DTI’s strength lies in acohesive approach to assisting customers.This cohesive approach means DTI workshand-in-hand with companies to ensure theyare getting the best performance, highestreliability, shortest time-to-market and themost efficient use of computing hardware fortheir program's embedded application. DTIhas a history of design and manufacturingexperience with standardized form factorssuch as PCI, ETX, COM Express, VME, CompactPCI and AdvancedTCA.

No part of this publication may be reproduced, stored in any retrieval system, or transmitted,in any form or by any means, electronics, mechanical photocopying, recording or otherwise,without the prior permission of Diversified Technology, Inc.

DTI Wired is available free of charge to qualified subscribers, by emailing [email protected]

To be removed from our mailing list, send a removal request via email to [email protected]

Other product and/or company names mentioned in this publication may be trademarks orregistered trademarks of their respective companies and are the sole property of their respective owners.

Copyright © 2009 Diversified Technology, Inc. - All Rights Reserved

Providing a Cohesive Approach to Embedded Solutions

Message from the PresidentPat Busby, President of Diversified Technology, Inc.

Given the wide range of business and economic challenges in today’s markets, it’s become vitallyimportant to have experience and staying power. Since 1945, the average recession lasted 10 months,but from 1919 to 1945, the average dragged on for 18 months. The last two recessions - from Marchto November 2001 and from July 1990 through March 1991 - were shorter than average, lasting onlyeight months each. This is the 5th recession during DTI’s 38 year history (1971), and the 8th forparent company Ergon (1954). Downturns are a normal part of an economic cycle, but they alsoprovide opportunities for those seasoned and primed to take advantage. Yes, some businesses will gounder, but others will emerge stronger. Those desiring to be among the latter need to be careful aboutmany factors, including supplier choices.

DTI is the supplier of choice to many customers for design, manufacturing and delivery of ruggedembedded computers for their solutions. DTI also provides complete integrated solutions for telecom,medical and military customers. Our CompactPCI and AdvancedTCA portfolio is transitioning well intomilitary applications. Along with our new technologies such as VME, rugged mobile power, RFtechnology, and energy inversion products, DTI will continue to invest in R&D to innovate and deliverproducts and value as a long term supplier.

In this issue of DTI Wired we offer direction and guidance on compute, switch fabric, rugged mobilepower, and wireless technologies integrated into products which improve our customer’s financial return.With the complexity of disparate environments, high performance processing, and balanced resourcesDTI provides solutions for long term expandability, economic value and operational maturity. I hope youfind this publication helps you in navigating through the technology and product acquisition in anever-changing environment. We look forward to serving you for the next 38 years, too.

A. Patrick Busby




Introduction

The growth in use of COTS (CommercialOff-the-Shelf) hardware and software acrossnumerous markets, most notably in thetelecommunications, government/ military, andcommercial sectors has led to discussions overthe benefits a COTS-based platform providesover a proprietary or completely customizedsolution. These benefits revolve aroundreduced deployment costs, easier maintenanceand support, the removal of vendor lock-in, andan increased ease of integration and scalability.

These aforementioned virtues of COTShardware have been well documented andproven to be true for all but the most severe ofcircumstances.

AdvancedTCA has seen an up swing in usagein classical military COTS recently. As a PICMGspecification (3.x), AdvancedTCA (ATCA)focuses on providing high speed interconnecttechnologies and improved reliability,serviceability, and availability to telecomapplications; many military applications demandthe exact same needs. Developed as an ‘Open’Standard (meaning PICMG member organiza-tions all have access to the specs todevelop their own boards, chassis, andadd-ons), ATCA seeks to provide a true COTSenvironment in the traditionally proprietarymilitary industry, and it is seeing growingsuccess in this mission due to many of theadvantages mentioned above that a COTSbased environment brings to a market.

There are, however, issues that can arise whenmoving to a COTS-based platform for yourorganization’s application. We will call theseissues “Gotchas” in this text, and explore notonly the financial impact, but also other nega-tive results on your organization’s developmentand deployment of your application that willoccur by you neglecting to investigate theseCOTS Gotchas.

COTS Gotchas come in two forms:

Feature Gotchas and Supplier Gotchas.

Feature Gotchas refer to claims made aboutthe specifications of a product as it compares tothe standard itself and as it will be used withinan application, essentially forcing you, as apotential COTS purchaser, “Does this productactually do what it claims to do?” or “Does thisproduct have all of the things the datasheet saysit does?”.

Supplier Gotchas, however, are more nefari-ous and insidious in their undermining of yourorganization’s success. They typically do notshow themselves until late in the game, typicallywell after a purchase order has been placed!Supplier Gotchas can be summarized as theend results of what a supplier promises versuswhat the supplier actually delivers. SupplierGotchas can be applied to nearly any COTS en-vironment and are not solely the property ofATCA suppliers. Any organization looking to

COTS GotchasWhat to Look for When Selecting a COTS Provider


move to any COTS platform should keep thesein the front of its mind. For this reason, thisarticle will focus primarily on Supplier Gotchas.

To note: This is by no means an attack on COTSmethodology or use. Commercial Off-the-Shelfhardware is certainly preferable in nearly everyway to proprietary deployments, and I do notintend this as an attack on COTS as a whole.The following is simply a discussion of key areasof interest that must be planned for and looked atclosely when selecting your COTS supplier.

Overview of the COTS Gotchas

In discussing COTS Supplier Gotchas, we mustaim to stay relevant. Therefore our COTSSupplier Gotchas will revolve around some ofthe most critical supplier services organizationsdesire when looking at a COTS Solution :

• Validation and Compliance

• Support and Maintenance

• Deployment Costs and Life Cycle Management- Obsolescence/EOL Planning- Asynchronous Upgrades

• Certification and Training

• Custom Hardware Design- Differentiation

• No Vendor Lock-in(Continued Open Standardization)

These six COTS Gotchas are by no means theonly considerations you must make whenlooking into a COTS Supplier, however they dorepresent some of the more important issuesthat can arise.

Before we move into explanations of the aboveGotchas and discussing the impact they canhave on your organization, we must firstunderstand that what follows is by no means animplication of COTS products as a negative.I fully support the use of COTS productswherever it makes sense to do so and firmlybelieve that the touted benefits of COTS are verymuch real and tangible.

These Gotchas are issues you must be aware ofwhen selecting a COTS provider.

Validation and Compliance

Validation

Validation refers to a solution being ‘validatedand tested’ with your application and in a field-like environment. One of the main benefits ofproprietary solutions is that due to a closerelationship with the provider, your organizationcan be assured that the solution will be validatedto integrate with and work in your specificapplication. The concern is that when dealingwith a COTS product, it has been only generallyvalidated within the scope of similar usage or,even, not validated within any ‘like’ application.This has been called Black Box Testing,meaning boards are released once basicapplication equivalency is demonstrated withinthe scope of the Supplier’s testing methods.

If a board passes testing, such as environmen-tal, or demonstrates like-application equivalencywithin the COTS Supplier’s testing facility butfails within your deployment, will or can theCOTS Supplier provide the support you need toovercome your unique issues, in say a bargetest?



I will put forth and state plainly the notion thatsome COTS Suppliers can provide this sort ofvalidation and testing, but you must choosewisely. Generally speaking, a more focusedCOTS Supplier will be more willing to demon-strate exact application usability with theirproduct rather than simple equivalence testing,and you can be assured that this COTS productwill not only work within your application, but thatthe COTS Supplier will indeed be willing, ready,and capable of helping you overcome anyunique problems that may arise.

COTS Gotcha Validation:

Select a COTS Supplier who can and will devotethe time necessary to solve the unique issues inyour deployment.

Compliance

Many organizations in the United States aretrying to achieve compliance to regulatorystandards and testing or to comply with non-American standards to supply their productoverseas . The most common standards beinglooked for are MilSpec 810F.

Many organizations will claim a product meetscompliancy. Very rarely, however, does asupplier actually take their product through thetesting. COTS products, when they claim aretypically “designed to meet” rather than haveactually passed testing. For your use, if yourproduct needs to meet specific compliancy, youwill have to take, and pay for, the qualificationsyourself.

COTS Gotcha Compliancy:

When selecting a product designed to meetspecific compliants, be sure that the supplier hasa history of taking products through that testing

and is intimately familiar with the process andtesting procedures. Tying in with Validation, besure the supplier will be willing to work very, veryclosely with you to help you pass your testing.

Support and Maintenance

Simple Support

Once you have deployed your solution withCOTS products, you will be required to maintainthe application throughout its lifecycle. Will yourCOTS provider be willing to give you in-depthsupport and maintenance help?

Many COTS providers will claim global supportand most of them will charge you for providingit. But what is it that these COTS providers areoffering? Part-timers following a flowchart withscripted replies, elevating a problem once criticalstatus has been reached, your application hasbeen down for a week (leaving you losing notonly revenue, but the respect of your end user),all the while your interaction with the Level 1support guys for that time has left you reduced tobegging and pleading to speak to someone whois familiar enough with the product and yourapplication to actually offer you some modicumof actual, real help.

Realistically the above will be the least of yourworries should you find yourself involved in“Waiting on Feedback” limbo. Even after Level1 phone support elevates your issue, will it beresolved? Many times, the next level of supportdoes not involve the COTS provider attemptingto solve your issue; rather it involves the COTSsupplier looking for any slight variation betweenintended and actual board use in yourdeployment in order to label your issue as ‘notmy problem’.


COTS Gotcha Simple Support:

Try to talk to other customers of the proposedCOTS supplier to get a feel for how they handlesupport requests. Also, evaluate just how closeof a relationship will be forged between you andyour COTS provider, similarly to how you shouldto make sure their Validation procedures willmatch your needs. You should focus onselecting a COTS supplier who will be dedicatedto solving your unique issues swiftly, rather thansimply focused on reading a script.

Deployment Costs & Life Cycle Management

One of the key benefits of utilizing COTSproducts is a lower deployment cost for yourapplication. While it is absolutely true that initialdeployment costs within a COTS platform arecheaper than a proprietary solution , there areassociated Gotchas when looking at the overalllifecycle of a platform, from initial deployment toproduct upgrades.

Obsolescence/End of Life Planning

One of the biggest concerns with using COTShardware is that of obsolescence and EOLPlanning . For our purposes, we can refer to theDepartment of Defense’s definitions of obsoles-cence and End of Life. Obsolescence isdiminished manufacturing and materialavailability of a COTS product, and End of Life(EOL) is a manufacturer’s notice to customers ofa last time buy of a product.

Obsolescence and End of Life occurrences aredefinite events that must be planned for withcare and exactitude when developing a COTSapplication. Ideally your organization willconduct market research on its own to survey

leading technologies, solid COTS productspresently available, and assess technologytrends to determine a plan for your productrefresh and upgraded COTS insertion. Includedin this would be a declaration of expected EOL ofa product from your preferred COTS vendors.

All products will someday be declared EOL bytheir manufacturer. Reasons may be that theirvendors have stopped producing the necessarymaterials, or an upgraded version of the currentproduct is now offered and the manufacturer willnot produce more of the older product due to thecosts. By having accurate and thorough EOLplanning, your organization will be prepared todeal with last-time buys of a dated technologyand be more ready and able to deploy a next-generation iteration of your application, whileavoiding many pitfalls and headaches.

By thorough obsolescence and EOL planning,your organization can have actual, real life cyclemanagement of your application when using aCOTS platform. It is not an impossible task – farfrom it. A more frustrating problem exists, andthis problem is not generally planned for: Lifecycle asymmetry. Every system has a life cycle. A system ismerely a set of parts, and each of theseindividual parts has its own life cycle, whichprobably will not coincide with one another. Upgrades will allow an extension of overallsystem life time by replacing obsolete or EOLparts of the system. But, given the nature ofdevelopment cycles and part availabilityschedules, component upgrades can becomeout of sync with each other.


Even in a simple hypothetical two-board COTSsystem, this can be demonstrated. If one boardhas a life cycle of 18 months while the 2nd is ona 24 month schedule, the next-generationrelease dates will only align once every sixyears. For every board you add on a differentschedule, your release alignment probabilitydecreases exponentially. This does not eveninclude patches, updates, firmware revisions,and other changes.

Asynchronous life cycles are likely to cause thegreatest headaches and problems of your COTSsystem deployments.

Your organization may find that although youhave accounted for your applications life cycleand part EOL situations, your vendor’s develop-ment lifecycle can cause you headaches when it

comes to upgrades. Sometimes the vendor’sdevelopment cycle happens to slip due to un-foreseen circumstances. Ether way, your appli-cation suffers and you have been hit by yetanother COTS Gotcha.

COTS Gotcha Lifecycle Management:

You must fully implement an obsolescence andlifecycle management program focusing onacquisition instead of problem resolution afterdeployment. Without this in place, a COTSsystem will cause more headaches down theline than it will solve. To this, add in the problemof Asynchronous Upgrades. You must plan forthe life cycles of the individual boards within yourCOTS system and plan for implementing thefuture technologies in your application withoutloss of uptime.

RUGGED COMPUTINGThe PlexSys-RP from DTI is a self-contained,rugged CompactPCI based portable computer.This MIL-STD chassis is equipped with 2 handlesfor easy transport making it a completely portablebattlefield workstation.

Multiple CPU Blade and Switch configurations areavailable including Intel’s Core 2 Duo, Dual-CoreXeon, and Low Voltage Pentium M.

Call NOW

for Immediate

Deployment

1-800-443-2667


Training/Skill Updating

When developing your new COTS application, itis likely that your engineers will be charting newwaters. They may be unfamiliar with the formfactor, fabric, or any number of othercomponents that will be used in your new build.In any case, they are going to have to becomeintimately familiar with the technology in order tosuccessfully develop and deploy your next-generation COTS-based application.

There are two ways in which your Engineers canlearn about this new technology. First, they canread the product documentation (manual) anduse their search engine of choice to finddiscussions, papers, and information relating tothe technology. I think it should be fairly obviousthat this would be a time consuming andwasteful method. To truly understand your newCOTS deployment, your Engineers will needtrue training and learning experiences that focuson your application.

Some COTS companies, if they are truly seriousabout their technology and YOUR success, willbe willing to invest the time needed to provideyou with evaluation and test units, give onlineseminars, answer specific questions viatelephone, and a rare few will be willing to holdworkshops at your lab or location to teach yourengineers the technology on a near one-to-onebasis. They will do everything in their power totrain your engineers in the effective and efficientuse of their technology and instruct you on howbest to implement their COTS solution withinyour application – not for their own gain butrather for your success.

So, when evaluating your potential COTSsuppliers, you must ask yourself, “How will myengineers be trained and gain the skillsnecessary to effectively implement this solutioninto our application?”

COTS Gotcha Training:

Try to find a COTS Supplier that is willing to forma cohesive bond to essentially become anextension of your engineering team to help in thetraining and skill development of your engineers.Doing so will give your engineers the ability tofully implement and maintain your COTSsolution.

Differentiation through Hardware?

You have probably been told that in a COTSmarketplace, your solution differentiation comesvia software. Let’s extrapolate in case youhaven’t:

Can software provide you with double or triple(or more) performance increases between twogenerations of your product? Very unlikely.These kinds of gains are only obtained throughhardware upgrades. If you and your competitorsare using the same COTS hardware you will notdifferentiate yourself from them in terms ofperformance – you and your competition willhave become aligned, attempting to outdo eachother via marketing instead of by out perform-ance.

Some COTS suppliers can and are willing toprovide not only minor “look and feel” changesbut also fully customized features, peripherals,and even boards for your COTS deployment.Sometimes non-recurring efforts are needed toassure future success through differentiation.



A good COTS provider will be able to give yourapplication the hardware performancedifferentiation that is needed to ensure yoursuccess in the marketplace. After all, to a goodCOTS supplier, your success is their success.

COTS Gotcha Differentiation:

While we’ve heard time and again that withCOTS deployments your differentiation mustoccur via software, you should determine thevalue of a COTS supplier who will also provideyou with the hardware differentiation you needand desire. If you fail to do this, the performancealignments among you and your competitors canmake it difficult to justify your end-solution toprospective customers.

Vendor Lock-In

One could make an argument that theabandonment of Vendor Lock In is a ‘Holy Grail’justification for the use of COTS in yourdeployment – which is a wonderful thing. Whenusing COTS, gone are the days of jacked-uphardware, integration, and support prices solelyfor the reason that the provider can charge what-ever they feel they can. That’s not to say thathardware, integration, and support prices can’tbe exorbitant, but with a COTS ecosystem themarket dictates pricing structures, rather than asingle company. COTS vendors have to com-pete not only on feature set, lead times, support,and capability, but also on price.

Oversight organizations within the COTSmarketplace help ensure interoperability amongmanufacturers to maintain an open marketplaceso users of COTS hardware are not forced topurchase from one vendor. If COTS suppliersmaintain adherence to standards and practices

within their market, you should encounter veryfew interoperability problems. But what abouttimes in which a COTS vendor chooses to takea proprietary route of some form while claimingfull adherence to an open marketplace and theinteroperability standards therein – somethinglike a proprietary switch stack?

Using the proprietary switch stack example,interoperability can become a problem. Want touse Switch 1 from Vendor A with a proprietarystack and Switch 2 from Vendor B with astandard stack in the same application? Youvery likely will run into severe interoperabilityproblems that will bring your development anddeployment schedule to a standstill until theproblem is solved.

COTS Gotcha Vendor Lock-In:

Just because an organization claims to be aCOTS supplier does not necessarily mean thattheir components will be fully interoperablewithin your specific application. You should, atthe very least, test the solution before selectingyour COTS supplier. If you have already chosenyour supplier and run into the aforementionedproblem, I hope you have chosen one of the fewsuppliers who will be ready, willing, and able todevote the time and resources to overcomingyour issues.

Key Fact for DTI:

One in every five AdvancedTCA Switches being

used today were designed and manufactured

by Diversified Technology, Inc.

Ref: VDC Report, 12/3/06


The above six items represent key areas thatyou must be aware of when developing anddeploying a COTS based system. If your COTSsupplier is lacking in any one of the aforemen-tioned areas, you can be sure that you will haveto overcome costly delays and headaches.

When planning your COTS system, you mustaccount for all areas of Lifecycle Costs, frominitial deployment to planning for obsolescence.Your COTS provider must be willing to work foryou to overcome your unique issues, essentiallymaking themselves an extension of yourengineering team. You must be aware ofvendor lock-in traps while selecting a COTSsupplier that will be capable and willing to giveyou the differentiation you need via hardware –not just passing the buck of true differentiationto your software developers. Also, your COTSprovider should be willing to train your engineersto get the most out of the hardware, be it viaemail, telephone, web-based training, or mostimportantly, on-site training.

If you incorporate heavy planning in all of theseareas to your COTS development plan, you cannot only be sure of real success, but also of aheadache-free deployment leading to a muchfaster time to revenue for your organization.

Article Written by:

Joe McDevittChief Technology Officer

Due to our increased focus on Military and Governmental applications, Diversified Technology, Inc. isconsidering bringing our computing and leading switch design expertise into the VME/VPX marketplace.If you have VME/VPX needs, or would like to discuss future plans therein with DTI, please contact Joe.

Joe McDevittChief Technology OfficerDiversified Technology, [email protected]

VME Solutions



Diversified Technology, Inc. realizes that standard products do not meet the needs of every application.To satisfy these design-specified requirements, DTI routinely develops custom solutions. These designscan be as simple as adding or removing one feature to an existing catalog product, or they may be ascomplex as a full platform design implementing next generation technology. In either case, thedevelopment cycle is simplified by DTI's in-house tools, capabilities, and manufacturing lines. AdvancedPCB and mechanical design software combined with libraries of function blocks streamline the initial de-velopment. BIOS source code engineers ensure the functional configuration is optimized to customerneeds. Rapid prototyping is achieved on DTI's state-of-the-art SMT assembly lines and in DTI's metalshop. Finally, DTI's verification and validation labs guaranty that the product is fully operational andmeets industry standards and can pass certification testing.

www.DiversifiedTechnology.com/custom

Elmer and his wife, Wendy, are the proud parents of two children; daughter, Arianna age 5 and Son,Elmer IV age 3. He and his family are active members of Triumph Church in Vicksburg, MS where heservers as a Church Elder, Board of Trustees Member, Teacher, and Former Sunday School Director.Elmer loves spending quality time with his family, playing golf and basketball.

Elmer currently serves as the Design Group Manager for DTI and isresponsible for the overseeing and coordinating of all hardware designengineers, the mitigation of program roadblocks, and assuring that projectmilestones and objectives are met. Elmer graduated from Mississippi StateUniversity with a BS in Electrical Engineering and a minor in Mathematics.He first joined Diversified Technology, Inc as a Hardware Engineer eightyears ago. In his role he enjoys debug efforts, solving tough problems, andgrowing others to be their best.

Custom DesignsApplication-Specific Computing Solutions

The Employees are DTI’s Most Important AssetEmployee Spotlight - Elmer Durrell III

All trademarks and tradenames are the property of their respective owners.


www.diversifiedtechnology.com

Diversified Technology, Inc. (DTI)

1.800.443.2667 • [email protected] • www.diversifiedtechnology.com/atca

Visit us on the web to find out more about DTI’s AdvancedTCA products.

ATS1936

Flexible 10 Gigabit

AdvancedTCA Switch

BOTTLENECKS SUCK

ELIMINATE CONGESTION WITH DTI’S 10 GIGABIT ATCA SWITCH


Traditionally in embedded hardware, specificallyin blade systems, the host or processing nodeslot has been the focal point of chassisfunctionality. The architecture itself lent the CPUnode to being the control, managing everythingfrom IP traffic to I/O connectivity. In Ad-vancedTCA, the core of the system is no longerthe CPU blade. Not only is the IP trafficprocessed at the switch level with Ad-vancedTCA, but also most of the I/O connectiv-ity will be processed over IP, creating a situationin which the switch drives the overall chassisfunctionality.

With the development of the next-generationbattlefield and network-centric warfare, many inthe military are finding that bottlenecks of an ap-plication’s functionality are not at the traditional

processing level but rather at the switch level,where the system cannot keep up with the band-width demanded. With ATCA, the switch drivesnot only the overall functionality of a givenchassis, but also the overall performance of theapplication.

Hesitation of ATCA Adoption

Bandwidth Increases from 10GbE

As the implementation of bandwidth-intensiveapplications continues to grow within theCommunications industry, the need for a drasticincrease in throughput becomes not onlyobvious, but absolutely necessary in order toprovide the kinds of feature-rich and robustbattlefield applications that the military requiresto remain successful in the evolution of the“connected battlefield.” There is no denying thatthe AdvancedTCA architecture is gainingtremendous ground within this sector, namelybecause of the increased processing power andbandwidth it provides.

The issue is, however, that althoughAdvancedTCA offers these increases inperformance, the rapid growth of bandwidth-intensive applications means that even thissoundness of architecture begins to choke andthe available processing power and bandwidthbecomes over-saturated. Here is where theCommunications industry must look forwardwithin the AdvancedTCA offerings to 10GEthernet fabrics.

10G EthernetAn AdvancedTCA Switching and Protocols Primer



Some target engineers have hesitated on theadoption of AdvancedTCA in the developmentand deployment of applications, mainly due tothe lack of bandwidth that 1Gb or 2x1GbEthernet provided in comparison to theirproprietary solutions. That argument is nowgone. AdvancedTCA board and systemmanufacturers are poised to offer the necessarybandwidth and reduced latency that telecommu-nications applications require (with room foroverhead), but can also offer the lower cost ofmoving to an open-standard architecture withwell-designed 10G solutions.

Ten-Gigabit Ethernet is the current leader(in terms of actual throughput capabilities) of theEthernet standards, providing ten times the datatransfer rate of the conventional (and most-used)1GbE solutions provided on most AdvancedTCAprocessing nodes and switches. Ethernet is afamiliar standard, and 10G is the naturalevolution, providing tremendously increasedthroughput and performance over current 1Gsolutions. Due to the low cost of implementationand corresponding astronomical use of Ethernetin nearly all data and communication

applications, its evolved form of 10G will be thedefault go-to for the Communications industry.Familiarity with the ‘core’ is key.

Ethernet is also a “best effort” fabric. Traffic isprocessed as best a switch can manage, but,unlike FedEx or UPS, there is no guaranteed“overnight” service. When a “best effort” fabric isin use, the best effort an application architect canmake is to over-design their AdvancedTCAplatform.

As stated, Ethernet is inexpensive to implement.If an organization were to over-design theirplanned 1G solution to 10G, they would find theadditional expenditure to do so is merely around4% of the overall chassis cost. Yes, four percent.This means that for a nearly indistinguishable

increase in cost, a solution can be over-designedto use 10G Ethernet, providing more assuredbandwidth within the platform.

DTI’s AdvancedTCA Manufacturing

DTI’s AdvancedTCA Testing


DTI’s ATS1936 is a flagship 10G ATCA Switchthat provides the bandwidth required on today’snetworked battlefield, as well as offering theflexibility to allow for direct customization throughAMC sites.

The ATS1936 is a true ATCA 3.1 Option 9 ATCASwitch, featuring 20-port 10G, wirespeednon-blocking fabric powered by Broadcom’sXGS3 silicon. With the same silicon, the baseswitch features 3 10G port, 23 auto-negotiating1G ports, and one 10/100 management port.

Ethernet management on the ATS1936 isprovided through a third party with emphasis onEnterprise support. DTI utilizes the Full SourceModel with the management software, whichruns directly out of user memory—negating anylicense exposure issues that are sometimesfound on other ATCA Switches. The Ethernetmanagement functionality fully supports IPv6routing, and offers superior performance incertain critical protocols, such as Spanning Tree.In utilizing the management software in thismanner, Diversified Technology has been able tofocus solely on providing superior ATCAfunctionality with the ATS1936.

The ATS1936 fully supports IPv4 and IPv6routing on both the Base and Fabric networks.The ATS1936 features 3 AMC mid-size sites, allof which support telecom linecards for telecomclocking options. The ATS1936 also supports anRTM with 10G Ethernet ingress/egress ports.Using a fully standardized design approach, DTIhas enabled the ATS1936 to provide the highestperformance at the lowest possible cost in anAdvancedTCA switch for use in telecommunica-tion applications such as VoIP, IMS, IPTV,Wireless, Security, and other solutions.

A Closer Look

Key Advanced Ethernet Features

The advanced Ethernet features of the ATS1936make it a truly robust and resilient solution forany military application, and provide the toolsnecessary to satisfy the aggressive need forbandwidth in today’s military applications.

Redundancy

The key theme in any AdvancedTCA systemand, truly, of the ATCA specification as a whole isredundancy. Within an Ethernet network, twokey protocols come into play when supporting

ATS1936 : DTI’s AdvancedTCA 10G Switch Blade with 3 AMCs



redundancy: Multiple Spanning Tree Protocol(an extension of Rapid Spanning Tree) and LinkAggregation Group.

Multiple Spanning Tree Protocol is an extensionof Rapid Spanning Tree and Spanning Tree.Spanning Tree Protocol (STP) allows loops andredundant links to exist on a single network byallowing only one active path between any twonetwork devices. Rapid STP works to make thenetwork adapt to changes much more quicklythan it would without it. Multiple STP is anextension of RSTP that gives VLAN support toSTP. With hardware that supports STP, RSTP,and MSTP, a network is provided 1+M (1 active,M Standbys) redundancy, because only one pathis allowed to be active at a given time.

In the presence of a Link Aggregation Group(LAG), network designers can combine multiplelinks between many network devices and groupthem as a single link, providing increasedthroughput and redundancy. By combining MSTP and LAG on the ATS1936,DTI has allowed for N-way active (all N active)redundancy for higher throughput andavailability between systems and networks.

Layer 3 Routing & Redundancy

When Layer 3 routing is used, Virtual RouterRedundancy Protocol (VRRP) works wonders ifthe switch is acting as a gateway. With VRRP,multiple routers can be set to appear as a singlerouter to nodes and other routers, which createshighly available gateways for Layer 3 networks.


If either switch goes down, the nodes willdynamically failover to use the other switch asthe network gateway. VRRP often does not addmuch cost to a switch. VRRP should beconsidered a requirement if Layer 3 Routing isto be used in a network, because it can be souseful.

Network Segmentation

Virtual Local Area Networks are a simple, butpowerful tool that increases security whiledecreasing broadcast traffic via segmenting thenetwork. While nearly every managed switchsupports basic VLANs, there are severalimportant extensions to VLANs that are notalways supported, including GARP VLANRegistration Protocol (GVRP) and protocolbased VLANs. GVRP can greatly ease thehassles of the configuration of larger networksby propagating VLANs automatically though thenetwork. Protocol-based VLANs allow fornetwork segmentation by protocols rather thanby physical ports. This is useful for networks thatneed to support diverse protocols.

Data Control

With high network traffic, important data shouldbe prioritized through what is generically calledQuality of Service. QoS can be implemented inmany different ways, but two of the more usefulmethods in ATCA are Access Control Lists(ACLs) and Differentiated Services (DiffServ),both of which are available on the ATS1936.

ACLs are simple rules that define how toprioritize traffic and are based on things such asIP addressing or a Layer-4 port number. DiffServ, on the other hand, allows for muchmore powerful rules, and in turn, better controlof data flow.

Multicasting with Layer 2 Networks

Internet Gateway Management ProtocolSnooping and GARP Multicast RegistrationProtocol bring the power of multicasting to Layer2 networks. Though multicast is not often used,it is critical to some networks. Without multicast,data is ether sent either to a single port or to allports of a switch.

Multicasting is a Layer 3/Layer 4 technology thatallows a router to easily send one set of data tomultiple recipients. Both IGMP Snooping andGMRP extend this technology to Layer 2networks. IGMP Snooping does so by“snooping” the higher level multicast packets andforwarding the packets only to the ports thathave members of the multicast group in thatpacket, whereas GMRP works by propagatingmulticast settings though a network. Both GMRPand IGMP Snooping reduce broadcast traffic onnetworks that use multicast.

Often unused, port mirroring can be a valuabletool for addressing network problems. With portmirroring all the data going out of a port, or ports,can be copied to another port for monitoring. Portmirroring can be used in traffic analysis, problemdebugging, and even security monitoring.

Remote Management

For large networks Remote Management is adefinite requirement. For AdvancedTCA, SNMP(Simple Network Management Protocol)provides the framework and is the key feature forremote management. CLI (Command LineInterface) does provide a management interface,but it is difficult to use programmatically. SNMPprovides remote management and monitoringthat is relatively easy to use in a programmaticway. Tools built around SNMP allow for constant


monitoring, statistics collection, dynamicreconfiguration, firmware updates, configurationbackups, and more on the entire network.SNMPv3 is of particular interest because of thebroad security features it brings to SNMP. Theend result of SNMP support is less downtimebecause problems can be identified and solvedmore quickly.

Problem Analysis

Often unused, Port Mirroring can be a valuabletool for addressing network problems. With portmirroring all the data going out of a port, or ports,can be copied to another port for monitoring. Portmirroring can be used in traffic analysis, problemdebugging, and even security monitoring.

Cost-Benefit of Advanced Ethernet

A Visual Cost-Benefit Analysis of Key Features

Every AdvancedTCA system has differentrequirements of the Ethernet switches used.Often the cost-benefit analysis of a feature willshow that the feature is needed, even when thecost is high.

Below are examples of features that are veryimportant to certain networks while not used inothers. The chart will show that many of the keyfeatures discussed above that are present on theATS1936 are of the utmost importance, but areonly negligible in their impact on the overall costof the system.

GO AHEAD AND JUMP INGO AHEAD AND JUMP INSimplify with COTS

Small Form Factor

Building Blocks

Small Form FactorsETX and COM Express CPU Modules

Call Today to Get your System

“Application Customized”

1-800-443-2667

Bringing it Together

What it all means

As stated earlier, with AdvancedTCA, the core ofthe system is no longer the host or system slothousing a processing blade, but rather falls tothe capabilities of the switch used in a system toprovide the bandwidth and features needed toadequately support an organization’s keyapplications. The rapid growth of the network-centric battlefield means that even thissoundness of the architecture of AdvancedTCAbegins to choke and the available processingpower and bandwidth becomes over-saturated.

Ten Gigabit Ethernet offerings are what thegovernment and military must look to in order tosolve these issues with their next-generationapplications, and the ATS1936 10G ATCA Switchprovides not only the necessary bandwidth butalso the key features needed on an advancedEthernet network. Even in solutions that only“require” 1G bandwidth, over-engineering to 10Gis an affordable and best effort design whenimplementing a fabric that takes a best effortapproach to bandwidth.

Article written by:JP LandryNetworking Program Manager



www.DTIRFSolutions.com

2MHz to 2GHz

SFRT Antenna Development

DTI also focuses on highperformance RF antennas forcustomer-specific applications:

- Enhanced performance forspecific bands

- Reduced coverage gaps

- Extended range

- Options for power limitedplatforms

- Reduced Physical andRadar Signature


Many embedded applications today require notonly significant processing performance, but alsoincreased I/O function in a compact form factor:Enter COM Express. In utilizing the COMmethod of a module containing all of thecomponents for a host computer (offered as anCOTS component) attached to a carrier board(COTS-developed or custom per application),the COM Express form factor is poised tobecome the de facto architecture for balancingperformance, size, I/O and thermals inembedded computing markets. By hardeningthrough extending operational temperatureranges and adding physical strengtheningcharacteristics, COM Express is also makingdirect inroads to governmental and military/aerospace applications.

In physical applications, COM Express can beutilized as a standalone embedded solution oras a processor mezzanine. To see this, we mustlook closer at the COM Express design standardand how its specifications allow it to be utilizedwithin an application setting.

Function and Design

The standard for a COM Express carrier board isCPU-agnostic. The architecture allows theembedded industry to adopt LVDS interfaceswhile maintaining legacy support.

The board-to-board connectors in COM Expressconsist of two rows, in which one row providespins for PCI Express, LPC Bus, SATA, LVDSLCD Channel, VGA and TV-out, LAN, system

and power management, and power and groundinterfaces, while the second provides SDVO andlegacy IDE and PCI signals with additional PCIExpress, LAN, and power and ground signals.

The maximum total I/O support in COM Expressbecomes:32 PCI Express lanes (80Gbps aggregate)

x16 PCI Express graphics2 LVDS Channels2 Serial DVO Channels4 SATA-150 links (600MBps aggregate)3 10/100/1000 Ethernet ports (10G provisionsin the future)8 USB 2.0 ports

COM Express has also been designed tosupport future generations of PCI Express andSATA interfaces.

The physical dimension requirements of theCOM Express CPU modules allows for twodifferent sizes: Basic and Extended, measuring125mm x 95mm and 155mm x 110mm,respectively. The Extended Form Factor has alarger power budget, doubles memory capacity,utilizes dual-channel RAM configurations, andallows the use of larger processors and chipsets.

The Basic and Extended Form Factors,however, share the same board-to-boardconnections, signaling definitions, and mechan-ical assemblies. This means that any carrierboard can support either a Basic or an ExtendedCPU Module.

COM-onomicsThe Economics and Use of COM Express in Embedded Applications


As can be seen, COM Express is a very flexiblearchitecture that can be used in a diverse rangeof applications, due to the significant I/Ocapabilities, CPU-agnostic design, and theability to interchange host CPU modules. So,how do these capabilities contribute to promotethe use of COM Express as either a standaloneembedded board or as a processor mezzanine?

COM Express – The Little Board that Could

Standards such as AdvancedTCA are focusedon a much defined market segment (Telco andData) because of three main reasons: First,significant bandwidth processing power. Second, there is no need, in most cases, forsignificant amounts of I/O within ATCAapplications. Third is the fact that, in theapplications that utilize AdvancedTCA, size isnot a very important issue. It makes sense, inthese cases, to utilize a separate processingnode, separate switch nodes, and to createexpansion via AMC modules. These threepoints can also be applied to MicroTCA, withlesser implications on necessary size.AdvancedTCA and MicroTCA are high-band-width, high-power, low I/O solutions mainly forTelco and Data applications. They suit theirmarket very, very well.

COM Express, however, is more flexible. Anyapplication in any market that requires highbandwidth, high I/O, and small size can utilizeCOM Express. Military/Aerospace, IndustrialAutomation, Test & Measurement, Security,Transportation, and Medical markets all requirethe processing power, I/O capabilities, and smallform factor that is present in COM Express.

As a standalone embedded board, the COMExpress methodology creates a solution in whichthere is significant processing and I/O bandwidthin a small physical footprint. Many applicationsin the aforementioned markets can takeadvantage of these characteristics by utilizing astandalone embedded board: Interactive kiosks,Self-Checkouts, training simulators, electronicbillboards and such can take advantage of thefeatures of COM Express by utilizing a stand-alone embedded board within their applications.

Within the same markets and applications,designers of embedded systems and applica-tions may also partition COM Express CPUModules from their own platform base, includingspecialty I/Os and FPGAs, utilizing COMExpress as a processing mezzanine. In doingso, these designers can focus on their applica-tion-specific I/O on a custom baseboard, withoutthe need to address high-speed switched fab-rics, signal reflections and impedance controls,routing, termination placement, etc.

The COM Express methodology creates anarchitecture that can satisfy the needs of most(if not all) applications that require a balance ofperformance and I/O in a compact form factor,and the flexibility of the specification allows forthe utilization of COM Express as either a stand-alone embedded board or as a processingmezzanine within the proprietary building blocksof other applications, allowing designers to focuson their own core engineering competencies.

While not an inherently rugged form factor,efforts are being made across vendors to offerextended temperature variations of theirofferings, as well as adding options such as



physical hardening characteristics for use inembedded military applications, such asvehicular controls, robotics, and simulatorcontrol mechanisms.

All of this is well and good, but the real questionsare “How will the adoption COM Express affectthe overall bottom-line of an organization?”What is the total cost of ownership, the productlife cycle costs, time-to-market, transitionalmetrics, etc.?”

Dollars and Cents

First and foremost, COM Express represents asignificant reduction in life cycle costs due to thescalability and interoperability of modules. Oncea specific carrier board has been implementedto an application, future hardware upgrades canbe implemented by simply changing the COMExpress Module. This significantly reduces thehardware costs of upgrading, which representthe bulk of expense when a new generation ofCPU and RAM technology becomes available.

In hand with the quick implementation ofupgraded hardware comes the faster time-to-market for deployment and time-to-revenueafter. There is a direct relationship between howfast an organization can deploy and implementhardware upgrades and how quickly saidorganization can begin to see ROI. With COMExpress, the essential plug-and-play architec-ture of the standard means that next-generationprocessor and RAM technology can bedeployed, implemented, and working muchfaster than previously, quickly generating arevenue stream. Simply stated, the COMmethodology allows an organization to rapidly

respond to demand fluctuations, competitiveforces, and new technologies by quickly andefficiently modifying existing designs, inexpen-sively broadening their product portfolios, andplug-and-play implementation of new COMboards.

COM Express represents a safeguard to R&Dinvestments and a lowering of total ownershipcost, because it allows designers to separatecommodity COM Express modules from theirproprietary building blocks. Investments intoR&D for specific, value-added applicationelements will not have to be reengineered forsuccessive technology generations.

It is possible to achieve some of the form-fit-function of COM Express via mini-ITX/nano-ITXmotherboards and the like at a lower cost. However, when dealing with these competingform factors, an organization loses significantlymore than it gains, mainly in the realm of lifecycle availability and revision control. COMExpress manufacturers strive to provide at leastfive years of availability and revision control oftheir products, allowing critical applications to beprovided with the necessary hardware through-out their lifespan.

Key Fact for DTI:

All of the Products designed and manufactured

by Diversified Technology, Inc. are done so in

the United States of America. We do not

outsource overseas in an effort to cut costs.


COM Express improves form-fit-function inspecific applications, and helps to reduce designrisks by cutting development time and costs,and, at the same time, providing the scalabilityand flexibility needed over a multi-year life cycleof an embedded application.

Finally, being a PICMG embedded standard,COM Express has no ‘proprietary’ builds. Standards-based COTS products have alreadybeen shown to reduce initial investment costs,as well as life cycle maintenance costs, due tothe inherent competition among suppliers toproduce cost-effective solutions.

As of now, COM Express represents the highestperformance available in small form factormodules. The scalability, flexibility, throughput,support for legacy and non-legacy standards,and low cost of ownership make COM Expressa very viable choice for a host of embeddedapplications across many different markets. The

ability to harden COM Express by extendingoperational temperature ranges and addingphysical strengthening characteristics is alsoextending COM Express into the realms ofgovernment and military/aerospace applications.

By reducing life cycle costs, protecting R&Dinvestments, and reducing time-to-market andtime-to-revenue metrics for any organization,COM Express makes a very compellingargument for its economy within the embeddedindustry.

Article written by:

Doug Mays, Field Application Engineer

CMXLX45 : DTI’s COM Express Module with Custom Baseboard

Baseboard

CMXLX45 Module



Current threats on our military forces havecreated a tremendous requirement for mobilityas it relates to mission specifications. Theobjectives of OBVP are to develop, demonstrateand transition electronic technologies thatenable lighter, physically smaller, more efficientand more reliable electronics, which are easierto operate and maintain. This in turn will enablecurrent and emerging electronic systems toincrease performance, have improved physicalcharacteristics, be more reliable and requiresignificantly less lifecycle resources. Furtherdevelopment of these programs will alsomaximize the advantages of increased DCsupply voltage inherent in the HMMWV, FMTVand HEMTT for delivery of a power solution formultiple military platforms in excess of today’savailable 10KW.

Sometimes OBVP is referred to as "Under theHood Power", but whatever the vernacular, itsimply refers to providing an alternative sourceof AC power on the battlefield that is generatedfrom the vehicle itself, as opposed to hauling alarge, heavy, diesel generator around on a trailer.What's the big deal with this concept you mayask. Well the military has been haulinggenerators around for decades and quitesuccessfully. When the enemy was confrontedat well defined lines and occupied a specifictheater of operation, that may have worked; buteven then it was just barely adequate enough toaccomplish the mission.

So what's changed? The truth is the U.S. hascome through a long dormant period, nearly 30years following the end of the cold war, wherewe as a nation got complacent while the eventsin the world were changing rapidly around us. Indefense of our military, the need to change wasrecognized much sooner by them, than we as anation through our elected officials. An apatheticview of our vulnerability developed that hasresulted in our current state of military readiness.The events of 9/11 and in Iraq were a shock andan unfortunate wake-up-call. Since that timethere has been a major shift in military strategy.The primary emphasis is no longer on watchingour borders for approaching missiles orbombers. We now are faced with a new enemy,terrorists, who scatter like roaches, hide in thedark and pop up anywhere and everywhere on aglobal basis.

Our military is now in a rapid transformation tobecome a lighter, more agile, mobile and anextremely more lethal force. To accomplish thismany technical and logistic problems must besolved; however, simply throwing money at aproblem doesn't always produce results. Eventoday the old engineering adage, "theprobability of obtaining an answer to anundefined problem, is zero" still applies.

Within the realm of this "military transformation"problem definition there are opposing, ordivergent forces, i.e., to become more lethal andbetter networked involves new weapons

Mobile Power for Mobile ForcesOn-Board Vehicle Power


systems and more sophisticated electronicsystems, both of which demand more DC andAC power. How do you get more power?...ahyes...larger, heavier generators. That makes the"lighter" and "more agile" components of the"transformation" equation harder to achieve.

The demand for fuel on the battlefield hasdoubled in the past ten years and the demandfor tactical power has quadrupled during thatsame time period.

Should that be a surprise to us? Well no. Isn'tit logical that one might conclude that thedemands of the war effort have resulted inadvanced technology, in the form of sophisti-cated electronic systems, improved weaponssystems, state-of-the-art field medical equipmentand elaborate networked Tactical OperationsCenters (TOC's). These all put a strain on thepower systems currently deployable to thebattlefield. In looking forward at what might beanother 20 to 40 years of anti-terrorist, militaryengagement, its easy to say that the fuel andpower requirements on the battlefield willexperience the same levels of increase over thenext ten years.

So how does this monumental problem getsolved? The best way is to break it up intosmaller pieces. Identify the chunks that can beaddressed with today's technology and begin todeploy, not all, but some portion to the battlefield.Then as technology allows, focus on tomorrow’ssolutions.

The first thing that can be done to lead totomorrow’s solutions is to focus current effortson the largest installed base of wheeledvehicles, that being the different variants of

HMMWV's in operation by the joint militaryservices. The large majority of the rolling stockof HMMWV's today contain a 28vdc electricalsystem.

True, there is a move to possibly increase this to42vdc, 100+ vdc and beyond in new-to-marketmodels currently being designed and tested.However, a nice, clean, simple retrofit couldresult in an upgrade of the existing fleet toinclude On-Board Vehicle Power in the range of8 to 10 KW. While if we had our druthers wewould want 30KW, it has been estimated thatbetween 70 and 80 percent of the dieselgenerators in the military inventory today are inthe 10KW power range. So addressing thissegment of the HMMWV installed base providesan interim step on the way to the more powerfulsystems planned for tomorrow.

By providing the HMMWV with a 7 to 8 KWOn-Board-Vehicle-Power system supplyingcontinuous AC power, the need to trailer a dieselgenerator is reduced significantly. In addition,now the HMMWV can negotiate terrain that itcould not previously engage while towing agenerator thus leading back to the more agileconcept.

What kind of AC power is required? Since themissions vary between the branches of serviceand even within the same branch of service, theAC power requirements vary considerably. Inmany cases a simple conversion from thevehicle's 28vdc source to single phase 120vac,60hz is adequate. In that regard, a number ofcompanies are providing this invertertechnology; however, the demands for power onthe battlefield, more often than not, exceed thislevel of sophistication. Therefore, the challenge



is designing an On-Board Vehicle Power systemthat meets multiple mission demands,regardless of the branch of service. This is by nomeans a minor task. There are simply timeswhen the little ole lady from the televisioncommercials says, "Where's the beef?", i.e.,120/208, 3 phase VAC.

Extreme Power Requirements in the Military

Now, regardless of whether the soldier isneeding single phase 120 vac or 3 phase, 208vac, in Europe where 50hz is prevalent, or if he'son the flight line, or a radar installation wherepower at 400hz is the frequency of choice, theOBVP system should be able to provide all of thefrequency options. With sophisticated commu-nications, computers and other specialtyelectronic systems, Total Harmonic Distortion(THD) becomes important since harmonic distortion emitted by the AC power source canrender this type of equipment inoperable. Therefore, a low THD is essential to completingthe desired operation.

As one reviews the specifications for an OBVPsystem, to be installed on 28 vdc vehicles, thereare certain performance minimums that becomeevident, as it relates to the Output Power, OutputVoltage, Input Voltage, Total Harmonic Distortionand Output Frequency. Physical characteristicsand mechanical dimensions tend to vary fromvendor to vendor such as length, width, height,weight and efficiency (heat dissipation) of thechassis.

Listing of parameter requirements:

Power Level

10kw peak,7 Kw continuous

Output Voltage

120/208 vac ( single phase and 3 phase )

Input Voltage

24 - 28 vdc

THD

Less than 5%

Output Frequency

50,60 or 400 hz (selectable)

Dimensions

Small as possible

Weight

Light as possible

Efficiency

High as possible

Among the major challenges in developing a 10KW OBVP system, especially a 208 vac, 3phase version, from a vehicles 28 vdc electricalsystem, is the internal power dissipation of theconverter itself, without creating a heat problemwithin the enclosure. Therefore, cooling is amajor challenge, especially if the goal is toconductive cool the system so as not to requireforced air through fans.

To obtain the necessary 28 vdc, 400 amps theHMMWV must either already be equipped with a400 amp alternator or be retrofitted with one.While most HMMWV's today are equipped with200 amp alternators, this is easily resolved sincethe Army has 400 amp, retrofittable, alternatorsin their inventory. Once the 28 vdc, 400 ampalternator power source is installed inconjunction with the 10 KW OBVP PowerConverter system, there is the problem ofkeeping the alternator running at an rpm that isadequate to supply the current to the PowerConverter system, which in turn will satisfy theAC electrical load, as well as, provide a continu-ous positive charge to the array of batteries. Toaccomplish this, an engine speed control systemmust become an integral part of the overallOBVP solution by increasing the engine rpm'sas the output load demand increases. Likewise,as the load is removed the engine rpm's mustreturn to idle.

Once the alternator, battery package and speedcontrol have been defined the actual designconcept or method of converting the 28 vdc to120/208 vac (single phase or 3 phase), 50, 60,or 400 hz selectable power can be addressed.Some methods of solving the dc to ac conver-sion involve dc-link or high-frequency ac-linktechnology. The most common form of switch-mode dc to ac conversion is characterized by adc link; however, the use of the dc-link haslimitations imposed by the dc-link itself. Anothermethod involves high-frequency ac-links thatdirectly convert a higher frequency ac voltage toa lower frequency ac voltage. This involves therectification and impulse excitation of a low-passfilter with integral half-cycle sinusoidal or quasi-sinusoidal, voltages originating from a source ofhigh-frequency ac voltage. While both represent

solutions, the optimum solution to the 28 vdc tohigh ac voltage, multi-phase requirements oftoday's OBVP system involves a variant of theseexisting technologies.

Through a series of events, that spanned anumber of years of research in the field ofOn-Board Vehicle Power, Diversified Technologyhas introduced an OBVP system concept thatmeets the specifications addressed in thisarticle. The system includes a power conversionelectronics chassis that is shock-mounted insidea ruggedized, waterproof enclosure.

Power Conversion Electronics Unit, VPS-10K

Description of OBVP Application

The original system definition for DTI’s OBVPprogram targeted the creation of compact,mobile power generation to be used for theSentinel Radar System in stationary modeapplications. As our development continued,additional applications were realized due to thesystem’s functionality and multiple mountingplatforms. Mobile power became even more ofa desire therefore the ability to produce poweron-the-move was incorporated in to the unit.




More urgent applications arose for OBVP withthe recent aftermaths of hurricane Katrina and2008’s Gustav. These include the capability ofsupplying power to ravished areas in dire needas it relates to first responders, health carefacilities, water purification centers, communica-tions centers, gas / fueling centers, etc. Usingthe capabilities of the OBVP system, instantpower could be realized in any area where aHMMWV could maneuver. National Guard firstresponders would have the tools necessary toprovide instant relief to those needing it most.

More and more programs require additionalpower generation capabilities. Future develop-ment of the OBVP program will target higherpower generation capabilities. Using the 10KWsystem as the framework, higher power levelscan be achieved by utilizing additional DCgenerating techniques for tactical wheeledvehicles. The focus is to make this technologyavailable for already established vehiclescurrently in use as well as for new programsunder development as part of the vehicleintegration package to power more advancedtechnology for immediate requirements in thewar against terrorism.

The development work for DTI’s programfocused all its efforts in providing a new solution,never before seen by the armed forces, andsupplying a modular approach of power genera-tion in a smaller, smarter and durable concept.This method of power generation provides thecapability of achieving power levels only seen inthe past by the utilization of a trailer towedapproach. The compact nature of this designallows for installation that does not affect cargospace needed for mission requirements. Beingwaterproof and not trailer towed, the HMMWV

now can operate under preferred missionrequirements without having to modify itsmission due to limitations of equipment deploy-ment. This provides the soldier with moreoptions in accomplishing objectives withouthaving to jeopardize safety based on limitsimposed by the previous solutions. Flexibility,mobility, portability and durability were the mainpoints of design during development. With theVPS-10K from DTI, those objectives areachieved.

The military hierarchy is moving to become alighter, more agile, mobile, networked, lethalforce through a modernization effort within thePentagon, known as "transformation". Theprofile of our enemy has changed from the moretraditional adversaries to those threats imposedby the far more elusive terrorist movement that isglobal in nature.

This article is an effort to divide the technologicalchallenges of providing tactical On-BoardVehicle Power to the battlefield into levels ofdifficulty. Thereby resulting in a "real-timesolution" for the 10kw and under systems, whileacknowledging that there is intense researchbeing conducted to attain the higher powerdemands of the future.

Article written by:Barrie McArthurVice Chairman Special Projects

MOBILE POWERFor Mobile Forces

www.DTIRuggedPower.com

What is OBVP? OBVP provides 3-Phase AC Power for Rugged Environments

OBVP offers Electronic Power that is:

• Solid State Electronics (No Moving Parts)• Physically Lighter• Smaller Dimensions• Easily Mountable • More Efficient and More Reliable

• Easier to Operate and Maintain

- DTI’s VPS-10K is Now Being Deployed -

On-Board Vehicle Power (OBVP)

Three Phase and Single Phase Operation Availablefor Mobile or Stationary Power

•••••

Hassle Free, Trailer Free7KW Continuous, 10KW Peak Power

•••••

50Hz, 60Hz and 400Hz Selectablefrom One Power Inverter

•••••

Minimizes Logistic Space RequirementsAboard Aircraft or Watercraft

•••••

Simple Retro-fit for Vehicle Integrationand Multiple Mounting Options

•••••

Provides Power to Teams, Patrols, Convoys,

During Unexpected Delays

•••••

Offers Simplified Fuel Logistics forMission Planning and Deployment

•••••

Back-up Power for Mission-Critical Equipment


Defining Load Balancing Simply

Load Balancing is a fairly simple concept: work that would normally be performed by one system is dis-tributed amongst many systems based on some predetermined criteria in a transparent fashion such thatthe client or end user is unaware of the separation of services. This is not what is traditionally thoughtof as Clustering, although Load Balancing is usually a feature of clusters. Typically load balancing isused by Internet service provides to distribute the computational and bandwidth loads of applicationsacross many servers in a server farm.

In this type of setup, the load balancer distributes requests to the backend servers, who reply to the loadbalancer. The load balancer is the “point of contact” to the client. The client or user cannot contact andis unaware of the presence of the backend servers and the separation of the services amongst them:

Basic Load Balancing Diagram

Layer 4 and Load Balancing

When the term “Load Balancing” is used, what is typically meant is Layer 4/7 Load Balancing. Thisrefers to the fourth and seventh layers of the OSI Model of networking abstraction, called the TransportLayer and Application Layer, respectively. In the OSI Model, the Transport Layer is where we find TCPand UDP protocols, segmentation of the data stream, error correction, and flow control. TCP (Trans-mission Control Protocol) provides, among other things, reliable end-to-end communication (is “con-nection-oriented”) via virtual circuits, error detection and congestion avoidance at Layer 4, whereasUDP (User/Universal Datagram Protocol) does not, and therefore has a much lower overhead than TCP.UDP is used at Layer 4 for application services that do not require guaranteed delivery of datagrams.

The Cohesive ApproachHow to Achieve High Availability Load Balancing with HAProxy on ATCA


TCP and UDP use port numbers to keep track of multiple application connections to the same location,essentially keeping track of connections to higher level protocols. What are called “well-known” portsare reserved for 1 – 1023. For example, TCP uses port 80 when initializing HTTP connections to aserver, port(s) 20/21 for FTP connections, and UDP uses port 69 for TFTP (Trivial FTP) connections.Some protocols may use both TCP and UDP to form a connection, but will only use one port, such asDNS, which uses both TCP and UDP to transmit data but only uses port 53.

The Application Layer (Layer 7) interfaces directly with and defines application processes and services,such as FTP, TFTP, HTTP, DNS, DHCP, SSH, SIP, and TELNET. The Application layer should not beconfused with desktop or user applications – these applications, such as a web browser, use theprocesses and services of Layer 7, but they do not reside in this layer themselves.

Layer 4/7 load balancing involves making distribution decisions based on the port number of an in-coming connection, which indicates the application level process or service that is using the transportlevel protocol. For example, one may choose to load balance incoming connections to a web server.In this case, the load balancer would be told something akin to “incoming connections to port 21, orFTP connections, should be distributed among Servers 1, 2, and 3.” The load balancer then does justthat, transparently balancing the bandwidth and computational load of the user connection.

Basic Load Balancer Distribution to Backend Application Servers

There are various methods of connection scheduling within a load balancing implementation. The mostbasic, and most common, is called Round Robin Load Balancing (which is what is pictured above). Ifa fifth connection were to be made above, that connection would go to Server 1, a sixth would go toServer 2, and so on. The connections are distributed in a round-robin fashion among the four applica-tion servers. Some load balancing solutions may use destination or source hashing, shortest expecteddelay scheduling, least-connection scheduling, or any of a variety of other methods for scheduling thedistribution of connections.



AdvancedTCA

What is AdvancedTCA?

For high availability and high bandwidth needs in telecommunications, commercial and industrial sec-tors, and in government/military applications, there is little doubt that AdvancedTCA (ATCA) presents oneof the most robust and powerful architectures available. Combining the performance and high availabilitywith the ease of scalability and upgrades, ATCA should be a strong contender for most applications insuch sectors.

For the uninitiated, AdvancedTCA is the largest specification effort in the history of the PCI IndustrialComputer Manufacturer’s Group (PICMG) and is targeted to requirements for next-generation com-munications equipment, and AdvancedTCA is finding homes in military, commercial, and industrial ap-plications. This series of specifications incorporates the latest trends in high speed interconnecttechnologies, next-generation processors, and improved reliability, high availability, and ease of serv-iceability and focuses on key benefits of a COTS (Commercial Off The Shelf) methodology.

ATCA was created with six key goals in mind:

• Enable interoperability among manufacturers’ components• High Availability (.99999 Uptime)• Reduce Time to Market/Revenue• Reduce development costs• Reduce the length of development cycles• Create an open, standards-based scalable, flexible platform architecture

As of today, ATCA has achieved these goals. In using the ATCA architecture, organizations will achievenot only a demonstrably reduced Time to Revenue, but lower overall development and deploymentcosts compared to traditional 1U or 2U deployments and proprietary solutions. Interoperability is a hall-mark of ATCA, allowing the choice of best-in-breed components for any application – preventing lock-in with one vendor whose proprietary solution may not meet all of an application’s unique needs.

As a bladed architecture, ATCA offers many benefits over traditional 1U and 2U solutions, namely in re-duced integration and cabling costs, as well as being extraordinarily scalable and flexible. A shelf canbe adapted to fulfill nearly any need via AMC expansion sites on most blades (or on a carrier blade) orthrough specialized full-sized blades fitted for one specific purpose.

The achieved interoperability allows the selection of the best components to fit the needs of an appli-cation. One vendor’s CPU node will work just fine with another’s switch in a third vendor’s chassis.ATCA offers the freedom of choice, the freedom of flexibility, and the freedom from vendor-enforced up-grades.



AdvancedTCA Switching & Networking Basics

An ATCA system is designed around a redundant dual star topology of communication among bladesin a chassis. A typical 5-Slot chassis will house two switches and three computing nodes, as well as aShelf Manager (ShMM) for system-level management. The blades may have AMC or RTM capabilitiesfor board-level I/O expansion and to allow for customization of board and system capabilities.

A 5-Slot chassis fully loaded resembles the following:

5 Slot ATCA Chassis with 2 ShMMs,2 Switches, and 3 CPU Nodes

As stated, the intra-chassis communication links are routed through the chassis backplane among theboards in a redundant, dual-star topology. Since each blade has a “base” and a “fabric” interconnect,each CPU node has two connections to each switch – one base interface and one fabric interface. TheBase interconnect is treated as a “control plane”, while the Fabric interconnect is called the “data plane.”Some ATCA backplanes support a full mesh architecture, in which every node (any endpoint) is con-nected to every other node. It is still possible to implement a simple dual star, but the expanded con-nectivity is available.

Redundant Dual Star Topology


In the above diagram, the red 1GbE connections are the base interconnects, while the blue 10G con-nections represent the fabric interconnects. All of the shown connectivity is intra-chassis, meaning it isrouted through the backplane of the chassis itself; there is no external cabling needed to connect thenodes and switches to one another. As one may guess, the above diagram, when implemented “naked,”would result in a whole host of networking problems like loops, congestion, and broadcast storms. How-ever, ATCA switches support a great many protocols to solve these issues, such as M/R/STP, FlowControl, VLANs, QoS/CoS, etc.

Representation of Full Mesh Topology

Most ATCA switches are Layer 2 switches and support hundreds of protocols. This allows for clean, ef-ficient, and effective networking at the Data Link layer of the OSI model and aides in developing flat net-work design. Some ATCA switches, like DTI’s ATS1936, support many Layer 3 (Network Layer) routingfunctions like RIP/v2, OSPF/v2, static routing, VLAN routing, VRRP, etc. While we still refer to these asLayer 2 switches, they do have greatly expanded and robust routing functionality.

Even fewer switches support the aforementioned Layer 4 functionality that Load Balancing requires.That is not to say they do not exist – one could find a very few ATCA switches that have Layer 4 LoadBalancing functionality integrated. However, these switches will typically break the bank when it comesto affordability!



The Set Up

Choosing the Basic Solution

Assuming an existing application utilizing AdvancedTCA with a switch that does not inherently supportLayer 4 load balancing, how could one take advantage of the nature of ATCA and the COTS method-ology to deploy a Load Balancing solution?

There are three basic methods (in terms of hardware) that could be implemented. First, buy an exter-nal load balancing appliance and force it as the connection between the external and ATCA networks.Secondly, since most ATCA system resources are under-utilized, if there exists a blade that isn’t doingmuch, or an extra slot in the chassis where a new blade could be inserted, we can develop a Linux-based, High-Availability Load Balancing proxy server. Lastly, we could replace our switch with one thatsupports Layer 4 Load Balancing, which seems to be an awful lot of money and an awful lot of time spentall for just one feature!

Let’s assume that we have a 5 Slot chassis with Diversified Technology’s ATS1160 switch and two DTIMPCBL0030 CPU nodes. The two CPU nodes are LAMPs – one is a live web server and the other isour backup. We would like to implement a Layer 4 Load Balancing solution that would make both ex-isting blades live servers that split the cost of bandwidth and usage, as our website has seen a drasticincrease in bandwidth and computational usage, maybe because our marketing and advertising de-partment has launched a new campaign focused on driving traffic to our website, and it has been wildlysuccessful.

Since we can get another MPCBL0030 for about one-half to one-third of the cost of a Layer 4 switch,we decide that the second method mentioned above (developing a high availability load balancing proxyserver) is the best method, in terms of a stop-gap, until we have the time and the budget to fully evalu-ate and deploy a true, hardware-based load balancing solution. Once our board arrives, we put it in ourchassis and prepare to develop our load balancing solution:

Targa-5 with 3 MPCBL0030 Nodes and 1 ATS1160 Switch


The Load Balancing Software (HAProxy)

For our load balancer, we’re going to be setting up HAProxy (http://haproxy.1wt.eu/), which focuses onhigh availability load balancing and automates failover on TCP and HTTP-based applications.

The documented featureset is as follows:• Able to route HTTP requests depending on statically assigned cookies • Spread the load among several servers while assuring server persistence through the use ofHTTP cookies

• Switch to backup servers in the event a main one fails • Accept connections to special ports dedicated to service monitoring • Stop accepting connections without breaking existing ones• Add/modify/delete HTTP headers both ways • Block requests matching a particular pattern• Hold clients to the right application server depending on application cookies• Report detailed status as HTML pages to authenticated users from an URI interceptedfrom the application.

It seems HAProxy will meet our needs quite nicely, and offers a few very beneficial features, such asreporting, Layer 7 persistence, and automated failover and recovery.

The Operating System

First we must take stock of where we stand with our current deployment. Web Server 1 (WS1) is ourlive server, Web Server 2 (WS2) is our backup, and the new board will be our High Availability Load Bal-ancer (HALB).

HALB needs an OS, and in this case we’re using Linux. I like CentOS 5.2, so that’s what we’re goingto go with. Since CentOS is a derivative of RHEL, the methods of deployment on RHx or Fedora shouldbe very similar. Other distributions can do what we want, but we’re going for enterprise-level high avail-ability, so CentOS it is.

Serial Redirection and OS Install

We will be running and operating HALB via serial console, so we need to install CentOS 5.2 via serialconsole, as well. A key word of advice: if you are going to be interacting with the board via serial redi-rection, install your OS via serial redirection – it will make your life much easier. In order to begin weneed an RJ45 cable and a rolled-null serial adapter, which we will connect at one end to the console portof HALB (the one marked IOIO), and the serial end to our laptop’s serial port. Open your favorite ter-minal program (I use TeraTerm), and set it to connect via COMx (whichever COM is the one with the se-rial adapter connected). We need to set our baud rate to 115200, data to 8 bits, no parity, 1 stop bit,

no flow control (115200, 8n1).



Baud Rate: 115200Data: 8 bitsParity: NoneStop: 1 bitFlow Control: None

Now, we could boot/install via network if we desired, but let’s just say (to make things simple) that wehave an external USB-based DVD-ROM drive and a CentOS 5.2 DVD. We connect this to the USB portof HALB, insert our CentOS DVD, and engage HALB with the chassis (shutting the Hot Swap handlecompletely). Once the Shelf Manager gives the ‘OK’, HALB will boot and we’ll get output through theconsole into our terminal program.

If no text is present past POST, it may be necessary to enter the BIOS (press DEL on POST) and en-able console/serial redirection.

At the CentOS install screen, there are installation options and the ability to input boot parameters.Since we’re installing via serial console, we must do a text install, and there is a very important boot pa-rameter we must use: console=ttyS0,<baud rate>. This must be used as a boot parameter, and mustbe placed before the word “text” when installing in text mode:

*********************************************** Press Enter to install graphically* Type linux text to install in text mode* Press the Function Keys for more options* Enter boot parameters below**********************************************boot: linux console=ttyS0,115200 text

Installation is fairly straightforward from here out. There are a few choices for packages to install, and Irecommend not installing the Virtualization or Clustering packages unless you really, really need them.The Virtualization package is simply Linux Virtual Server and Xen, which have been integrated into thekernel, and the Clustering package is OpenMosix and a few other applications. It is always possible toinstall these at a later time should they be necessary.

When asked to set up networking, eth0 should be the default active port, assuming your connection tothe external network is through the base interface of your switch, rather than through a front panel Eth-ernet port on the MPCBL0030. If your external connection is through the fabric channel of the switch,eth3 should be the default. Set your IP address and other parameters as required.


Once the OS has installed and you have rebooted (it will force reboot in order to complete setup), I liketo immediately update and upgrade everything.

To Note: While CentOS is an RPM based distribution, if you’d rather use APT than Yum, the processto do so is:

1. Download the proper release package for your hardware from:http://dag.wieers.com/packages/rpmforge-release/

2. Install the package: [root@HALB ~]# rpm –i rpmforge-release-0.3.6-1.el4.rf.*.rpm

3. Now perform[root@HALB ~]# yum install apt

4. Now update /etc/apt/sources.list.d/os.list and add the following lines:***CentOS Entries***repomd http://mirror.centos.org/ centos/$(VERSION)/os/$(ARCH)repomd http://mirror.centos.org/ centos/$(VERSION)/updates/$(ARCH)repomd http://mirror.centos.org/ centos/$(VERSION)/extras/$(ARCH)

5. Now you can use apt-get. Try it out[root@HALB ~]# apt-get update[root@HALB ~]# apt-get upgrade

HAProxy Installation and Setup

Remaining House Cleaning

As stated HAProxy is a high availability TCP/HTTP load balancer. Since the concept application is anHTTP/80 load balancer, we’re going to need a free Ethernet interface. What is meant by this is that wecannot have another service bound to the main Ethernet interface (such as Apache) or the TCP portwe’re going to use of it’s IP address.

So, there are still a few things to do before tackling installing HAProxy. First, we need to stop Apache(if it is running) and make sure it will not return on system restart:



[root@HALB ~]# service httpd stopStopping httpd: [ OK ][root@HALB ~]# [root@HALB ~]# chkconfig httpd off

Apache has been stopped and will not initialize if there is a system reboot. It should be noted that wewant to leave Apache up and running on WS1 and WS2.

Sometimes there are issues with Kudzu, the automatic hardware probing mechanism in RHEL/Cen-tOS. I generally turn it off at this point, as I won’t be adding any hardware to the board in question.

[root@HALB ~]# chkconfig kudzu off

HAProxy Archive and Installation

We need to download the HAProxy executable:

[root@HALB ~]# cd ../usr/src[root@HALB ~]# wget http://haproxy.1wt.eu/download/1.2/src/haproxy-1.2.18.tar.gz

--13:50:22--http://haproxy.1wt.eu/download/1.2/src/haproxy-1.2.18.tar.gzResolving haproxy.1wt.eu... 88.191.22.98Connecting to haproxy.1wt.eu|88.191.22.98|:80... connected.HTTP request sent, awaiting response... 200 OKLength: 206219 (201K) [application/x-tar]Saving to: `haproxy-1.2.18.tar.gz'

100%[=======================================>] 206,219 15.3K/s in 11s

13:50:33 (18.9 KB/s) - `haproxy-1.2.18.tar.gz' saved [206219/206219]

Now we untar, make, and set it as a service to start on boot:

[root@HALB src]# tar -zxvf haproxy-1.2.18.tar.gz[root@HALB src]# cd haproxy-1.2.18[root@HALB haproxy-1.2.18]# make target=linux26[root@HALB haproxy-1.2.18]# cp haproxy usr/sbin/haproxy[root@HALB haproxy-1.2.18]# wget http://www.dtims.com/haproxy/ haproxy.init -O /etc/init.d/haproxy[root@HALB haproxy-1.2.18]# chmod +x /etc/init.d/haproxy[root@HALB haproxy-1.2.18]# chkconfig haproxy on[root@HALB haproxy-1.2.18]#

HAProxy Configuration

HAProxy should be generally ready to go. All that remains is to edit the configuration file to meet theneeds of our concept application and to actually start the service.

The HAProxy configuration file is stored at etc/haproxy.cfg

If the file does not exist, you can download an existing configuration file from Diversified Technology,Inc.’s website at

http://www.diversifiedtechnology.com/haproxy/haproxy.cfg

Overviews of HAProxy’s settings and configuration parameters are in the configuration documentation:

http://haproxy.1wt.eu/download/1.3/doc/configuration.txt

Within the HAProxy configuration file, there are 3 master sections: Global, Defaults, and Listen. Listedbelow are the values in the default configuration file on DTI’s website, with explanations for settings:

globallog 127.0.0.1 local0 info Logging

maxconn 4096 Maximum Number of connections on the front end

daemon Background process forking, allows nbproc parameter

nbproc 4 Number of processing cores on the blade. The MPCBL0030 has2 dual core processors, making 4 cores total.

user root

#debug Commented. Enable for debugging

#quiet Commented. Enable for debugging

defaultslog global enables per instance logging of events and traffic

mode http set running mode or protocol

option httplog enable logging of HTTP request, session state, and timers

option dontlognull disable logging of null connections

retries 10

option redispatch enables session redistribution if there is connection failure.requires a non-zero value for retries parameter.

maxconn 2000 maximum per-process number of concurrent connections

contimeout 5000 maximum time to wait for a connection attempt to a server to succeed

clitimeout 50000 maximum client side inactivity timesrvtimeout 50000 maximum inactivity time on the server side


listenlisten webfarm <IP of HALB eth3>:80 tells HAProxy to bind to port 80 of the eth3 IP address

mode http set running mode or protocol

balance roundrobin set the methodology of load balancing

server WS1 <IP of WS1 eth3>:80 weight 1 these two arguments tell HAProxy where to send thetraffic and in what proportion (weight)

server WS2 <IP of WS2 eth3>:80 weight 1

Here is where we need the IP address of HALB, WS1 and WS2. Since this is AdvancedTCA, we haveabout 6 Ethernet ports on each board to choose from! If we’ve architected our chassis properly, how-ever, eth0 or eth3 should be the only ports whose IP addresses are of concern, presently. If our serversare going to be serving via the Base channel, eth0 is our port, and if we’re serving on fabric, eth3 shouldbe our port. Let’s assume we’re serving on fabric—we need the IP addresses of eth3 on HALB, onWS1, and on WS2. Let’s imagine our current addressing looks like:

HALB:eth3 10.10.1.2 netmask 255.255.254.0WS1: eth3 10.10.1.3 netmask 255.255.254.0WS2: eth3 10.10.1.4 netmask 255.255.254.0

Given this addressing scheme, in the listen section of the haproxy.cfg file, our arguments that requireIP addresses will be as follows:

listen webfarm 10.10.1.2:80...server WS1 10.10.1.3:80 weight 1server WS2 10.10.1.4:80 weight 1

Once completed, save the configuration and start HAProxy:

[root@HALB ~]# service haproxy startStarting haproxy: [ OK ]

Now all that is left to do is to reconfigure your settings such that your URL points to the eth3 interfaceof HALB instead of WS1. In doing so, we have constructed a TCP/HTTP round-robin load balancing so-lution in relatively short order.




Our final setup resembles the following diagram:

Conclusion

HAProxy is a very simple and lightweight tool that can be used to quickly deploy TCP/HTTP load bal-ancing solutions. While it can be deployed on nearly any hardware, AdvancedTCA can present a prob-lem for those not very familiar with ATCA networking and configuration. HAProxy offers far more options,arguments, and parameters than those discussed in the configuration file above – it is far deeper thanI expected at first glance, and I’m sure there are many who have far more advanced configurations thanthe simple one I have put together above.

While the above setup is simple, it will work and work well, costing far less than purchasing dedicatedload balancing appliances or new ATCA switches with Layer 4 Load Balancing integrated into the stack.While a hardware solution will be more reliable in the long run, HAProxy is, by its very name, made forhigh availability uptime and should be a very good holdover until such time that a hardware solution canbe deployed.

Article Written by:Doug MaysField Application Engineer


Joe Speakby Joe McDevitt, Chief Technology Officer

Many internal and external people have asked me why I ran for office to be PICMG’s VP of Marketingand I can’t help but to think of the movie “A Beautiful Mind” staring Russell Crowe as John Nash, amathematical genius, eventual Nobel Prize winner, and sufferer of paranoid schizophrenia. Late in hiscollege career, Nash grows increasingly frustrated as he seeks in vain for his revolutionary idea. Nash’s“A-Ha” moment — as the movie purports — happens one night while drinking beers with his fellowmathematician students. While enjoying tall, frosted beers, five girls enter the bar. The sole blonde istaller and more striking than the other four, and the mathematicians begin to discuss strategies forsecuring a dance with the taller girl. They conclude that Adam Smith is correct and everyone needs towork in one’s self interest, leaving the prettiest girl to decide who is the fittest. Nash ponders thatapproach for a moment and then explains that if they all attempt to dance with the striking blonde, theywill block each other’s efforts and no one will succeed. Further more, the other girls will reject theirsecondary efforts, as no one wants to be the “second fiddle.” In the movie Nash concludes that theyshould all agree to completely ignore the blonde and approach the other girls first. Nash states thatAdam Smith’s theory is incomplete and that success comes not from everyone doing what is best forthemselves alone, but in doing what is best for themselves and the group. PICMG is about thosecompetitive developers gathering and agreeing to work together, and thereby attracting customers intoan ecosystem - much like Nash stated and is the main reason why I ran.

About Diversified Technology, Inc.

Diversified Technology, Inc. (DTI) is an embedded hardware company whose strength lies in thecohesive approach we use with our customers. This cohesive approach means DTI works hand-in-handwith customers to ensure they are getting the best performance, highest reliability, shortesttime-to-market and the most efficient use of computing hardware for their program's embeddedapplication.

DTI, an Ergon Company, was founded in 1971 and has a history of design experience with PICMGstandards. DTI has been prolific in introducing technology and standard off-the-shelf products based onPCI/ISA, COM Express, CompactPCI, AdvancedTCA and other standardized form factors. In additionto COTS solutions, DTI is also a leader in the custom development of advanced computing boards andembedded systems for unique, application-specific designs. DTI single board and system products arepointed toward mission critical applications such as telecommunications, factory automation,networking, ruggedized military, high-end gaming and medical control operations.


MS National Guard

Natchez, MSMay 1 to 2, 2009

NGB Logistics

Keystone, COMay 4 to 8, 2009

WINDPOWER 2009

Chicago, ILMay 4 to 7, 2009

RTECC

Greenbelt, MDMay 5, 2009

RTECC

Boston, MAMay 7, 2009

Border Security

Phoenix, AZMay 13 to 14, 2009

RTECC

Phoenix, AZMay 19, 2009

RTECC

Fort Huachuca, AZMay 21, 2009

Interop

Las Vegas, NVMay 19 to 21, 2009

Fires Conference

Ft. Sills, OKJune 1 to 4, 2009

RTECC

Montreal, QCJune 2, 2009

RTECC

Ottawa, ONJune 4, 2009

MS Chiefs of Police

Biloxi, MSJune 8 to 11, 2009

National Sheriff Association

Ft. Lauderdale, FLJune 22 to 23, 2009

SMD Conference

Huntsville, ALAugust 17 to 20, 2009

DSEi 2009

London, UKSeptember 8 to 11, 2009

NGAUS Conference

Nashville, TNSeptember 11 to 13, 2009

RTECC

San Diego, CASeptember 29, 2009

Modern Day Marine

Quantico, VASeptember 29 to October 1, 2009

RTECC

Long Beach, CAOctober 1, 2009

DTI Conference Schedule


*DTI’s Product Roadmap is subject to change. Please contact Diversified Technology, Inc. for a current revision.

[email protected] | 1.800.443.2667 | 476 Highland Colony Parkway, Ridgeland, MS 39157

DTI EMBEDDED ROADMAPP

ICM

G 1

.0

2009 2010 2011 2012 2013

Co

mp

actP

CI

AT

CA

VM

ES

mall F

orm

Facto

rs

LBC9426Dual-Core Xeon / Dual CPU

LBC9526Future Dual CPU/Chipset

LBC9626Future Dual CPU/Chipset

LBC9316Core 2 Duo

LBC9416Future CPU/Chipset

LBC9516Future CPU/Chipset

LBC9216Pentium 4 or Celeron

TrexSys Platforms1U, 2U and 4U Systems

CPB4612Pentium M

CPB4912Future CPU/Chipset

CPB4712Dual-Core Xeon

CPS4224Switch, 10G Egress (4),

19 Slot ChassisCPB4812Core 2 Duo

PlexSys Platforms1U, 2U, 4U Systems

ATC6231/ATC6239Dual and Quad-Core

ATC7239Future CPU/Chipset

ATC8239Future CPU/Chipset

ATS193610G Switch

ATS1930AMC (3) Carrier,

PCIe and 10 GigE

ATS7160AMC (2), Base GigE,

Fabric 10G Serial

CMXLX45COM Express Board

CMXLX55Future CPU/Chipset

ETXM690ETX Board

ETXLX15ETX Board

AMC & RTM BoardsApplication Specific ATI1914

AMC (1), 10G to 1GExpansion Board

Targa Platforms5 and 14 Slot Systems

VPX Switch

VPX CPUIntel Chipset/CPU

VPX CPUFuture Chipset/CPU

1.800.443.2667

www.Diversif iedTechnology.com


476 Highland Colony Parkway

Ridgeland, MS 39157

US PostagePAID

Jackson, MSPermit #134

Documents

DTI Custom Publication, 2009