Upload
iris-jacobs
View
214
Download
1
Embed Size (px)
Citation preview
Is Windows Right for High-Availability Enterprise Applications?Dan Kusnetzky, Vice PresidentSystem Software ResearchIDC
IDC © 2002
Agenda
What are “Enterprise Applications?”
IDC’s Spectrum of Scalability?
What is “High-Availability?”
What are the 7 meanings of the term “cluster?”
IDC’s spectrum of availability software
How the vendors stack up
Questions?
IDC © 2002
What are “Enterprise Applications?”
All organizations are “enterprises”
• Regardless of revenues or the number of
employees
• Vendors use the term “enterprise” to imply
things which may or many not be true
IDC © 2002
What are “Enterprise Applications?”
Questions to ask your suppliers
• Does this application or tool have a direct impact on:• Each customer?• Each employee?• Each partner or supplier?• If the answers are “no” then it’s not an
enterprise application
• Will the organization go out of business without this application?
IDC © 2002
IDC’s Model of Scalability
Complex/Small
Simple/Small
Complex/Large
Simple/Large
Complexity
Transactions/Day
IDC © 2002
VS.
Various Ways to ScaleMultifunction vs. Functional
Servers
Systems have defined functions
Database
Applications orapplication
components
Load-balancedWeb servers
Storage
IDC © 2002
What is “High Availability”?
Applications and Data remain available beyond
the life of its host
There are many ways to achieve this
High Availability solutions could include
• Application fail over
• Middleware
• Serverware
• Storage software
IDC © 2002
What’s a Cluster Among Friends?
The Goal: harnessing
the power of many
machines to create a
single virtual
environment
Each approach is
selected by different
people for different
needs
Parallel processing
Load balancing
High availability/fail
over
Single system image
Application fail over
Storage availability
and performance
IDC © 2002
Some “Clustering” History
VAXcluster and IBM’s Parallel Sysplex: high
water marks
Unix Clustering: Behind but catching up
Linux: evolving from Web load balancing and
HPTC to more commercial approaches
Storage Software Suppliers: Data availability and
application fail over
Microsoft – Taking a Different Approach
IDC © 2002
Load Balancing or Parallel Processing Monitor
Two to thousands of independent loosely-linked
systems Multiple systems have copies of applications
and data
Applications run on all systems
Monitor distributes workload among the available
systems
• Distribute loads using round-robin, request or capacity model
Data synchronization and administration can be
challenges
IDC © 2002
High Availability Monitor Two to 32 systems
cooperating to create a single environment.
Multiple systems have copies of applications and data.
Applications run on all systems.
• Communicate with a high availability monitor through special APIs.
• Data accessed through a parallel database or special APIs.
Monitor notifies systems
of an outage so
applications can respond.
Load balancing may not
be available.
Run applications in
parallel for improved
performance. Run multiple
copies of applications to
improve scalability,
Administration can be
challenge.
IDC © 2002
Clustering Monitor
Two to 32 systems tied tightly together.
Multiple systems have copies of applications and data.
Applications are run on all systems.
• Data can be accessed through a parallel database or directly as if on a single system.
• It may not be necessary to use special APIs.
Monitor notifies systems of an outage so the operating environment can respond.
Applications can be run in parallel for improved performance. Multiple copies of applications can be run to improve scalability.
Everyone sees a single virtual environment.
IDC © 2002
High Availability Applications
Two to thousands of systems cooperating to create a single environment at the application level
Multiple systems have copies of applications and data
Applications are run on all systems
Application contains
logic to handle failure
scenarios
Other applications
may not benefit
Administration can be
challenge
IDC © 2002
High Availability for Storage
Storage servers via NAS or SAN
Storage replication
Fail-over manager virtualizes storage
IDC © 2002
Microsoft’s Traditional Strategy
Own the the following, and you own the customer’s systems
• APIs
• Development tools
• File formats
• Communications architectures
Create incompatibilities drive customers to use only Microsoft products
Only Microsoft created “standards” are fully supported; others are not
IDC © 2002
Microsoft’s Approach to Clustering and High Availability Solutions
Philosophy
• Let our software do it – we know more about
your needs than you do
• Microsoft software everywhere, doing
everything
• Everything is legacy: should be encapsulated
and eventually replaced with a Windows
solution even if it working productively
IDC © 2002
Microsoft’s Approach to Clustering and High Availability Solutions
Layers
• Presentation (IIS, Site Server, SNA Server)
• Business logic (Application Center 2000,
COM+)
• Data access and storage (SQL Server 2000,
Windows 2000 now, Windows .NET Server in
the future, Microsoft cluster services)
IDC © 2002
How the Vendors Stack Up
• Microsoft Application Center 2000
• Red Hat High Availability Server
• TurboLinux Cluster Server
• Legato Cluster Server
• IBM HACMP
• Microsoft MSCS
• Mission Critical Linux Convolo
• Veritas Cluster Server
• Compaq TruCluster for VMS or TruCluster for Tru64 UNIX
• HP MC/Service Guard
• Sun Cluster 3.0
• Caldera/SCO Non-stop Cluster for UnixWare
• Open Source Beowulf, LVS and others
• TurboLinux EnFuzion
• Platform Computing LSF
• Sun Gridware
Parallel Processing Load Balancing High Availability Clustering Manager
Monitor Monitor
IDC © 2002
Clustering and High Availability Software Market Drivers
B2B, B2C and in-house applications can not appear to slow down or to fail
Staff with necessary skills are difficult to find and costly
Clustering and high availability software is:
• Difficult to install, configure and use today
• Will be much easier over time
Directed by operating environment adoption
Open Source alternatives limit potential for revenue growth
IDC © 2002
Is Windows Right for High-Availability Enterprise Applications?
Today’s Answer: A definite maybe.
Some applications are served well by highly distributed architectures
• Low intensity of interdependent data
• Algorithm allowing decomposition
Some applications are better when hosted on a single, medium or large scale system
• High intensity of interdependent data
• Monolithic application architecture
Tomorrow’s Answer: As the Eight Ball says “signs point to yes”
IDC © 2002
Related Research IDC#24798 - Clustering and High-Availability Software Market Forecast and
Analysis, 2001-2005
IDC#24844 - Linux Operating Environments Software Market Forecast and Analysis, 2001-2005
IDC#24827 - Windows Operating Environments Market Forecast and Analysis, 2001-2005
IDC#24799 - Web-Centric Computing Software Market Forecast and Analysis, 2001-2005
IDC#24846 - Server Storage Software Market Forecast and Analysis, 2001-2005
IDC#24851 - Unix Operating Environments Market Forecast and Analysis, 2001-2005