5 Steps to a Successful Relocation_1-2013

7/30/2019 5 Steps to a Successful Relocation_1-2013

1/8

1Smart move: Five StepStoa SucceSSFul Data center relocation

Smart Move:

Five Steps to a Successful

Data Center Relocation

As organizations relocate their data centers, many lack the up-rontplanning that reduces the short-term risk o business disruption andthe longer-term risk o choosing a acility poorly suited to theirongoing business needs.

By Tim Schutt


2/8


CROSS THE U.S., DATA CENTERS ARE LITERALLY IN MOTION. Manycompanies, rom manuacturers to retailers, rom health care to high tech, are

relocating their computer acilities in search o more ecient, eective and secureIT operations.

Today, data center operations represent about one-quarter o most companys total IT costs, somanaging relocation has taken on increased importance. Directing a complex data center

move is anything but easy. When done right, a move requires careul planningand expert execution. But, i poorly managed, a move can disrupt businessoperations or days, or weeks, once the new site goes live. Longer term, movingto a bad locationone with insucient power, suboptimal space, or aninfexible landlordcan orce a company to move its data center again longbeore it should.

Data center relocations require excellence in planning and execution. Commonmistakes such as ailure to properly evaluate the build versus lease decision, or relying too

much on an overburdened IT sta, can be prevented. In this white paper, we discuss typicalrelocation pitalls to avoid as well as the site selection, acility design, and relocation techniquesthat can make a data center move a success.

Today, more and more companies are planning to relocate their data centers. In act, accordingto a Gartner Group study, more than 70% o the Global 1000 organizations will have to eithermove or modiy their data center acilities signicantly during the next ve years. There are manyscenarios that justiy a data center relocation, the most common being the need to update aginginrastructure. The average age o todays data centers is 12 years, which requently leads to the

ollowing business and technical problems:

OUT OF POWERThe acility lacks sucient additional power required or upgrades to high-density server and storage systems. I 200 watts per square oot is required and only 100 watts isavailable, the only options are to retrot the space, use dierent systems or move the data center.

OUT OF COOLINGIn older acilities a point o diminishing returns can be reached where thecost o additional cooling increases at an accelerated rate. In this case, the only options are toretrot the space (assuming sucient power and airfow are available), consolidate systems ormove.

OUT OF SPACEWith the need or more power and cooling demanded by state-o-the-artequipment, operating a data center in traditional oce space is becoming increasingly dicult andexpensive. In some cases, the migration to high density systems is blocked simply on limitations inceiling height and the inability to introduce sucient cool air while adequately handling the heatedexhaust.

LEASE RENEWALSWith increased demands or power, cooling and space, a leased acilitymay be reaching the end o its useul lie. The acility owner may be unwilling, or economicallyunable, upgrade the acility to meet state o the art requirements or high density data centers.

The reasons why companies may choose to relocate their data centers are clear and compelling.However, since a data center move is oten a once in a career event, many companies experi-ence these same avoidable and costly mistakes:

ExEcutivE Summary

A

Why Data cEntErS movE

More than 70% o the Global 1000

organizations will have to either move

or modiy their data center acilities

signifcantly over the next fve years.


3/8


10 Years o Experience Shows Us That Data Center Relocations Go Wrong Because:

top 3 rEaSonS Data cEntEr rElocationS Fall Short


4/8


hoW notto managEa movE

thE right Wayto managEa movE

A number o companies have ound moving data centers to be anything but easy. For example,Hostway merged with Anity Internet to become one o the worlds largest Web hosting providers,

with 600,000 customers and 15 operations centers in 11 countries. Three months later, Hostwayplanned to move 3,700 servers, or 3,000 o its customers, rom its Miami center to Tampa, Fl. Thecompany warned customers to expect outages o between 12 to 15 hours during the move.

As it turned out, those outages lasted three days (and even longer or more than400 customers) resulting in lost revenue and prot, not to mention a serious blowto their reputation. Summarized by a Hostway vice president, The company sawan unusually large number o hardware ailures that occurred during thetransportation.

It literally sounds like Hostway took their servers, put them on a truck, unpacked them and tired toplug them in again, said one analyst1.

Another hosting provider supporting 165,000 websites planned to relocate 200 servers to a new

location. Customers were told their sites would be down or about 12 hours on a Saturday. In act,many sites were down or days. The truth is, relocating a data center is not a simple matter opulling out the plugs in site A and plugging them back in at site B, as many companies (to theirdistress) have discovered.

Fortunately, industry best practices do exist and can help guide a company through any datacenter move. The ollowing steps summarize the most important areas on which to ocus whenplanning a relocation.

STEP ONE:Build or Lease?The rst decision acing an organization that has decided to move is whether to build or lease.

When a company owns its own data center, it assumes greater maintenance costs, securityexpenses (such as guards), and other risks. Building a new data center rom scratch requires alarge capital outlay. For example, constructing a Tier 3 data center can cost as much as $10,000 to$20,000 per kW o capacity. Although building a new center gives an organization maximum fexi-bility and control, or most companies this option is too ar removed rom their core competenciesto be realistically considered. Relatively ew organizations have the expertise needed in real estateandconstruction, data center design, operations, security and maintenance.

On the other hand, leasing has its own challenges. I a company does not own its own acility, itslandlord can raise the rent or make other demands. We have seen landlords give 30 days notice

or renters to purchase more o its services or pay additional rental ees. In one case, rent wastripled and as a result, some tenants moved. Those unable to do so were seriously hurt by thisincrease in costs.

When leasing, long-term lease agreements are an absolute necessity. Relocation is simply tooexpensive, timeconsuming, and risky to be orced prematurely. This is especially true or largercompanies with bigger data centers. Companies must lease with the uture in mind. This meanscontracts should include extension provisions. It also means leases should contain expansionoptions or uture growth and possibly purchase clauses should the landlord alter nancially andbe orced to sell.

1Hostway Failed to Properly Plan Server Migration, eweek.com

Relocating a data center is not a simple

matter o pulling out the plugs in site A

and plugging them back in at site B.


5/8


Whether a company decides to build or lease, it cannot aord to trust to luck (or to any providersunvetted expertise) that the design o the data center will ulll its current and uture needs.Whether it builds or leases, a company can (and most oten should) outsource the design work i itlacks internal expertise in the design o state-o-the-art high-density data centers.

I a company is leasing a new data center, it is o course constrained in design by the availablespace and uture expansion rights. This will make optimizing the available space even more critical.At a minimum, the company must evaluate the ollowing key points and ask the right questions:

Capacity (Initial and Future)

Is there enough power and cooling? Are the power and cooling in balance or will one run out

before the other?

How can I reduce or eliminate expansion risk without overpaying now?

Is there enough physical space? When you need to grow, where will additional space be

provisioned (e.g., is it adjacent to initial space and on the same power / cooling systems)?

Facility Design Verifcation How can you verify capacity is as advertised for uptime design, redundancy, load, etc.?

Are there any failure points that could cascade through the facility?

Facility Maintenance

Are the systems properly maintained and are the records auditable for generators, air

handlers, pumps, chillers, uid levels, etc.?

Is maintenance outsourced or provided by facility staff? How are they hired and trained?

I a company is building a data center, it must be involved in all the design issues rom the start.These will range rom acilities concerns like veriying that foor-to-ceiling height is adequate, to themore arcane (yet critical) disciplines o rack conguration.

Regardless which path is most appropriatebuild or leasethe next step is to make sure thespecic site will meet your short term and long term objectives.

STEP TWO: Conduct a Site Suitability AnalysisA site suitability analysis should be conducted prior to either leasing or building a new data center.There are many actors to consider when choosing a site, including:

GEOGRAPHY. The data center should be located ar rom high-risk sites o potential naturaldisaster, such as foods, earthquakes and hurricanes. Locations near major highways and aircratfight corridors should be avoided as part o risk mitigation. The site should be on high ground, andprotected.

COMMUNICATIONS. The site should have multiple, ully diverse ber connections to networkservice providers.

ABUNDANT POWER. In 2010 data centers consumed about 2 percent o all electricity generatedin the U.S., and consumption is projected to continue its rapid growth (see chart next page, DataCenter Power Usage). Any site under consideration should have easy access to abundant powerrom multiple sources o electricity, taking advantage o low cost providers whenever possible.


6/8


SPACE. Data center relocations, as noted, are expensive, and one doesnt want to do them oten.Thereore, beore siting a new data center, the organization must analyze its planned businessgrowth. Dont build or lease based on your current needs -- always determine (as best you can)how much space and power youll need ve or ten years down the road.

Many other considerations related to location will also enter the picture. For example, youll needto consider the availability o water (needed as a reliable backup supply or cooling towers andchillers); environmental approvals or uel (including backup generators) and building exhaust; 24x7building security; and much more.

STEP THREE: Plan Thoroughly, and Well in AdvanceA successul data center relocation project requires comprehensive planning and preparation welloutside the scope o normal day-to-day IT operations. With many IT departments already stretched

thin, the chances or success greatly increase when outside experts are brought in to help. Whilethere are several ways to determine a move teams success ater the act, two o the best methodsare to measure system perormance and the satisaction o end users ater the move is completedBoth metrics should be veried to be equal to or better than they were prior to the move.

In creating a comprehensive move plan, the ollowing steps are recommended:

BENCHMARK APPLICATIONS. This will provide the basis or verication when the systems arereturned to service. Establishing a baseline also provides a validation point / rame o reerenceshould questions be raised post move regarding application perormance.

DETERMINE MAXIMUM ALLOWABLE DOWN TIME ACCEPTABLE FOR THE MOVE. Withproper planning, a move can be accomplished in a time window that is appropriate to the mission /

operations: rom zero downtime, to several hours or even a day or two. Forexample, high volume e-commerce businesses and applications where publicsaety is at risk may demand a no down time move whereas B2B operationsthat operate on a Monday to Friday schedule may be appropriate to move over aweekend. Some large-scale moves may make sense to accomplish in a series

o move events with each bundle being achieved in its own appropriate move window.

IDENTIFY SERVER, STORAGE AND APPLICATION DEPENDENCIES.As enterprise applica-tions cross departmental boundaries, it is important to dene unctional move bundles to assurethat ull business process support is returned to service as quickly as possible.

Any change undertaken during a move

adds risk and complicates the project.

Data Center Power Usage

Data centers account for 1.7 2.2% of all electricity generated in the U.S. today and are projected to rise in

the future despite the impact of virutalization and cloud computing.

SOURCES: ANALYTICS PRESS/STANFORD UNIVERSITY REPORT, AUGUST 2011


7/8


REVIEW ALL RELEVANT AGREEMENTS AND CONTRACTS. The list should include vendorleases, maintenance agreements, warranties and insurance policies. Be sure to determine howrequirements or restrictions may impact the moves success or introduce r isk.

IDENTIFY MOVE RESOURCES. What employees will be involved and what will be their role be

prior to, during and following the move? What supplier and service contractors will be involved?When multiple organizations are working together, make sure the chain o command andresponsibilities are clear.

CREATE A COMPLETELY ELABORATED MOVE DAY PLAN. (We call ours Runbooks). Justas NASA wouldnt try to launch without a complete fight plan with clearly dened owners or everyactivity, you need a thorough, minute-by-minute, plan or your move events. To minimize the risko making mistakes, your runbook needs to encompass everything rom application shutdown tosystem backups to physical move activities, and you need to stick to it on the day o the move.Automated tools can greatly ease the burden, here, too.

STEP FOUR:Minimize Complexity and Change

Prior to the move day, organizations should reeze their application congurations and notintroduce change and complexity. Sometimes, data center relocations are viewed as anopportunity to upgrade sotware, hardware, or network elements, as change seems easier toaccomplish in a fuid environment. This is a bad idea or many reasons. Among other issues,changing the application portolio at move time makes it dicult to measure the moves successbecause the pre-move baseline has been altered.

Any change undertaken during a move adds risk and complicates the project. This is especiallysignicant when it comes to todays popular practice o server virtualization. Resist the urge tovirtualize as part o a data center relocation. Virtualization is a signicant project in itsel, andattempting to virtualize servers during a move means trying to do two very dicult things at thesame timea recipe or disaster.

There is one exception to the no changes rule: An organization should consider the purchase onew core networking gear beore the move. This will signicantly reduce the risks and down-timeo reinstalling network gear at the new site during the move. The network should be up and ullytested beore any move is made.

STEP FIVE:Use Dedicated Resources on Move DayAs with all complex tasks, the devil is in the details. Managers undertaking a data center movemust create detailed checklists covering everything rom server rails to cabinets to specializedtools (e.g., screw threaders or stripped screws; sledge hammers; special power cords to preventoutages when servers cannot be reconnected, etc.). Extra manpower will accelerate the move,reduce down time and allow optimum use o individual skills.

Beore moving day, you will have already developed your timeline, manpower / resource plan,interdependencies and move bundles. By this time everything should be labeled (every device,every rail, every cabinet, every cable, everything), new cabling and network connectivity should beinstalled and tested in the destination data center, and trucking and manpower ready and waiting.

To limit downtime, a data center move should begin as soon as applications have been shut down,and the move should be scheduled at a time when involved employees are able to ully ocuson the moves success. Trying to squeeze in a move on top o ull time, daily responsibilities isdoomed to ailure. It will create a heavy tax on employee morale and unnecessarily prolong themove and potentially impact the business.


8/8


For example, one sotware company wanted to move 225 servers to a new data center. It plannedto accomplish this relocation over the course o ve weekends during the summer, eectivelyeliminating summer weekends or its IT sta. By reworking the plan and leveraging experts in datacenter relocation, this company completed the move in one weekend. This saved signicant timeand money, while keeping core IT sta ocused and motivated on day-to-day operations.

There is a truism in the military: No plan survives contact with the enemy. This means that nomatter how detailed the plan, the people who execute it must always be ready to make adjust-ments or the reality on the ground. Make sure the move day team knows the plan and theirresponsibilities. Make sure you have data center move experts on handthis is clearly not a goodtime or on-the-job training.

As IT becomes ever more essential to business success, companies are outgrowing their datacenters and data center relocations are moving to the center stage o IT execution. However,selecting a bad site or poor planning can result in higher expenses and another move ar too soon.

Worse, a botched move can stop an enterprise dead in its tracks.

Unortunately, operational IT knowledge does not translate into an understanding o how to bestmove a data center. Nor does real estate knowledge translate into the ability to select the best datacenter site.

Businesses need data centers that can both support their current operations and provide thefexibility or uture growth. As more and more companies contemplate data center relocation, theyneed to remember that a move is like a battle; it requires great planning, excellent execution andsucient manpower.

About the Author

Tim Schutt is a TDS Principal, and the Vice President o Data Center Relocation Services.Tim has over 25 years o experience managing people, processes and budgets as anexecutive and in operations roles at Palm, Inc., SunGard Securities, Paymentech and WangLaboratories.

About Transitional Data Services (TDS)

TDS provides independent assessments, recommendations and improvements or IT,including data center designs, relocations, operational support, ERP, and web and mobileapplications. Our recommendations cross departmental and technology silos to achieve thebest ROI or our clients. Since we do not operate as a vendor, VAR, or real estate brokerwe are unrestricted to a specic product portolio and unbiased by the latest trends and

highest commissions.

TDS clients include successul organizations o all sizes and industries including Kayak.com, The University o Texas, Boston Red Sox, Cedars-Sinai Medical Center, Liberty Mutu-al and many others. As these organizations are committed to leadership in their respectiveelds, they rely on quality business partners like TDS to operate transparently in their bestinterest.

Transitional Data Services. 2012 All rights reserved. 508.625.3030

www.transitionaldata.com

Why ExpEriEncE mattErS

Documents

5 Steps to a Successful Relocation_1-2013