Planning a Server Refresh q SERVER HARDWARE ROADMAPS q SERVER CONSOLIDATION q CAPACITY PLANNING q BENCHMARKING AND TESTING www.searchdatacenter.com
May 25, 2015
Planning a Server Refresh
q SERVER HARDWAREROADMAPS
q SERVER CONSOLIDATION
q CAPACITY PLANNING
q BENCHMARKINGAND TESTING
www.searchdatacenter.com
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
2 PLANNING A SERVER REFRE SH
Server Hardware Roadmaps
choosing a new server is hardly a random process.Wise IT administrators
take great care to evaluate prospective features, functionality, form factor and
facilities needs, and thenmatch these capabilities against the company’s com-
puting requirements.
But administrators and technology officers don’t stop there. In many cases,
server vendors must also explain their long-term product plans—the roadmap
—that their server hardware will follow for years to come. A server roadmap is
sometimes just as important as a vendor’s maintenance agreement, so let’s con-
sider the implications of server hardware roadmaps when planning a server
refresh.
IMPORTANCE OF THE SERVER ROADMAP
A server roadmap is the vendor’s attempt to predict the evolutionary path that
a server model or family will follow into the future. A roadmapwill often con-
sider technical specifications such as anticipated processor, memory, storage,
network and form factor support. As just one hypothetical example, upcoming
enterprise-class servers may support up to four AMD Interlagos 16-core
processors andmanaging 512 GB or RAM ormore. That translates to 64 proces-
sor cores on a single box, which can be quite attractive for highly consolidated
virtual servers.
Similarly, an organization might need to ensure that future form factors (such
as 1U or 2U) will fit existing physical rack structures, or they maymove to a
blade form factor for maximum processing in minimum space. There are also
storage considerations. For example, solid-state disk drive (SSDD) technology
has made inroads into blade and other small-form-factor servers, allowing local
caching and other performance improvements that might not have been possi-
ble with blades on a traditional storage area network (SAN).
SERVERHARDWAREROADMAPS
Although server virtualization can reduce total server count in a data center,
the use of more powerful systems (e.g., far more processors andmemory) does
not always carry a corresponding reduction in power and cooling demands. In
fact, the adoption of highly concentrated form factors such as blade servers can
profoundly affect cooling schemes. So server roadmaps are an important part
of facilities planning because organizations must ensure that adequate facilities
are in place before new servers are deployed.
“Wewanted to move to a blade-type system, but the [high-utilization] load
that we press onto our servers would basically melt the box,” said Chris Steffen,
principal technical architect at Kroll Factual Data in Loveland, Colo. He notes
that some new blade servers are specifically designed to accommodate heavy
utilization—a key example of roadmap planning.
But while the “specs” are important, experienced administrators dig even
deeper to consider factors such as warranties, maintenance, service-level agree-
ments (SLAs) and the continuing changes in virtualization support. Experts
note that the synergies between server manufacturers and virtualization
providers can be far more interesting and eye-opening than the server’s hard-
ware specifications.
OTHER FACTORS IN SERVER SELECTION
Organizations also face the challenge of choosing heterogeneous or homoge-
neous server platforms in a data center. Both approaches present pros and cons
for IT administrators. Perhaps the major argument for heterogeneous data cen-
ters is the reduced risk of vendor lock-in. “Wewere pretty big on trying to stan-
dardize on one vendor and on one model line,” said Ian Parker, a seniorWeb
services administrator at Thomson Reuters in AnnArbor, Mich. “But I’ve had
a lot of problems with OS and firmware and BIOS interactions.” Organizations
like Thomson Reuters now prefer a more heterogeneous approach: matching
the server to the computing needs with less regard for hardware models or
family lines.
3 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
SERVERHARDWAREROADMAPS
For other organizations, choosing a homogeneous server environment makes
strategic sense because it reduces documentation and eases the availability of
parts. It also makes financial sense because competitive vendors will typically
negotiate deep discounts for organizations that are savvy enough to commit to
the product line. “All things being equal, if I get to have a little bit more money
by having a complete Dell solution … then I’m going to take the money and
run,” Steffen said.
Other emerging technologies such as unified computing systems (UCS) have
little impact on server selection at this point. Experts express interest in the
semi-homogeneous premise but point out that the relatively small number of
players and the current cost-benefit proposition of the technology simply have
not generated much interest for large data center deployments. It is more likely
that UCSwill expand its role in small andmedium-sized deployments, where
the simplicity of a single-vendor model of servers, network and storage systems
makes more sense with smaller IT staffs and budgets.
Systemmanagement considerations should also be factored into a server
roadmap. “You definitely want to be able to manage stuff once it’s in the door,”
Steffen said. He added that it’s an important issue but not an all-or-nothing
decision because today’s management tools have evolved to the point where
they can usually provide adequate support for a broad cross-section of major
server families. Administrators should ensure that prospective server lines can
integrate with existing management tools either natively or through plug-ins,
PowerShell scripts and so on. The goal is to provide the deepest level of control
possible for future servers without having to adopt different or vendor-specific
tools.
STICKING WITH THE PRODUCT LINE
In order for an organization to commit to a server model or family, it’s critical
for a vendor to show a level of commitment to the product and the roadmap it
presents. This puts the vendor’s service and support high on the list of priori-
4 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
SERVERHARDWAREROADMAPS
ties for many businesses. Even though server virtualization allows server
workloads to be migrated or restarted quickly, most administrators still want
the hardware up and running.
A strong SLAwith favorable terms and pricing can go a long way toward
establishing andmaintaining customer loyalty. “You still need to have those
agreements in place, and they need to be competitive,” Steffen said. “Waiting
two weeks for them to turn around a new box is not going to cut it.”
Perhaps the most detrimental event for an organization is discovering that
a vendor has unexpectedly changed or
dropped a server roadmap, and a thought-
ful administrator will plan for that contin-
gency. Experts like Parker note that regular
communication from the vendor is impor-
tant. “One of the reasons I deal with major
vendors is because you’re going to commu-
nicate,” he said. “If I didn’t care about that,
I’d go buy white boxes from anyone else.”
But the rapid advance of technology will
inevitably throw speed bumps along the server roadmap. The product that you
expect to buy in three or five years may likely be replaced with an evenmore
capable and powerful product, which may be difficult—or even impossible—to
budget for adequately. So regular communication and updates from a vendor
can help organizations adjust to changes in the roadmap and plan for product
transitions in a more proactive way. �
5 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
SERVERHARDWAREROADMAPS
A thoughtful admin-istrator will plan forthe unexpected, suchas a sudden changein the vendor’s serverroadmap.
Server Consolidation
server virtualization and consolidation has become an integral part of an
organization’s technology refresh cycle. Administrators must purchase servers
that can support multiple workloads, provide extra computing resources to
handle workloads failed over from other servers, maintain reliable operation
over the long term and still fit adequately within the data center’s facilities,
including power, cooling and space constraints. Meeting this combination of
capabilities while staying within a reasonable budget can be a challenge for
busy IT administrators, but achieving a server consolidation strategy is a vital
part of server refresh planning.
THE QUICK-AND-DIRTY OF SERVER CONSOLIDATION
The concept andmotivation behind server consolidation are straightforward.
A traditional server runs one operating system and application, but only 5% to
10% of a server’s computing resources (CPU, memory and I/O) are actually
used. This means that 90% to 95% of a server’s computing potential—and the
capital used to buy it—is routinely wasted. By implementing a virtualization
platform, applications are converted into virtual machines (VMs), and the
physical server can simultaneously host multiple VMs as separate and distinct
instances.
The additional workloads use more of the server’s computing resources,
making muchmore efficient use of the box. Virtualization can easily allow 60%
utilization of the server, though some organizations may push a system to 90%
utilization. Because one physical server can runmultiple workloads, a data cen-
ter can reduce its total number of servers, which theoretically translates into
lower capital expenditures, reduced space requirements and lower operating
costs for power and cooling.
The requirements for server consolidation are surprisingly light; you’ll need
6 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
SERVERCONSOLIDATION
a hypervisor, servers, applications and storage. Examples of virtualization
platforms (or hypervisors) includeMicrosoft’s Hyper-V, VMware’s vCenter or
Citrix Systems’ XenServer. Organizations just getting started with server virtu-
alization can generally acquire a free version of each hypervisor for proof-of-
principle deployments. Physical servers must also be capable of running a
hypervisor, though virtually anymodern server within the current refresh
cycle should have no trouble.
Today, almost all professionally developed application software can run
properly as a VM, and organizations regularly virtualize Microsoft Exchange,
SQL and other high-end applications. Still, there is no guarantee that every
application will perform properly in your infrastructure, so it pays to test each
application in a lab environment before rolling the VM out into production.
Finally, each VM is basically a file that can be stored, updated, copied and
moved, so you should have a SAN in place to service each virtualized server.
SAN snapshots are a well-establishedmeans of backing up VMs.
PLANNING FOR SERVER CONSOLIDATION
Today’s hypervisors are quite refined and well-supported on a vast array of
server hardware, so there is no need to select a server specifically to run a virtu-
alization platform. But there are some collateral issues that will affect your data
center planning. Of course, the most important consideration is the choice of
new servers. New servers are rarely basic commodity boxes. They can be, but
it’s more likely that new servers will be large and powerful boxes with substan-
tial computing resources available. This allows the new server to host a large
number of VMswhile having the maximum impact on facilities needs.
Planning should also consider how older servers will be reallocated within
the organization. “If you have a good long-term plan, figure out how to use
those [older servers] somewhere else,” Steffen said. “Maybe that’s in disaster
recovery. Maybe that’s in a dev and [quality assurance] environment. Maybe it’s
something as simple as setting up a lab for some pet project.” In some cases,
7 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
SERVERCONSOLIDATION
older servers may be removed and sold off to the secondary market.
It’s also important to consider how virtualization affects other tasks within
the data center, such as data protection and backup. For example, VMs can eas-
ily be protected with snapshots saved to a SAN and then replicated to off-site
storage for disaster recovery. Other organizations may choose to stick with es-
tablished backup tools, but be sure those
hardware and software tools will run as
expected. Steffen recounts early prob-
lems introducing Hyper-V because the
older servers used for backups were not
64-bit machines. This required additional
servers to support the more demanding
software environment.
And there are other server consolida-
tion pitfalls that administrators must be
sensitive to. First, server consolidation should consider failover planning. If a
server hosting 10 VMs fails, all the VMs on that server will also become un-
available until the server is repaired. This means administrators must provide
a means to restart failed VMs on other servers with available computing re-
sources.
Virtualization also relies on SAN storage, so organizations will need to con-
sider the storage requirements and costs associated with their virtual servers.
“The more that you virtualize, the more that you’re looking at enterprise-grade
storage and the costs that are associated with that,” said Parker. “We find our-
selves spending more on storage—sometimes kind of aghast at the money we’re
spending.”
Another pitfall of server virtualization is VM sprawl, or the uncontrolled cre-
ation and deployment of newVMs. This happens because VMs can be created
in mere minutes, and organizations rarely implement the business policies and
procedures needed to control newVMs and their lifecycles. If left uncontrolled,
VM sprawl can easily overwhelm storage and available computing resources,
8 PLANN ING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
SERVERCONSOLIDATION
VMs can be createdin mere minutes, andorganizations rarelyimplement the businesspolicies needed tocontrol sprawl.
and this translates into much higher storage costs, management overhead and
license expenses.
MAKING SERVER CONSOLIDATION WORK
In general, server virtualization is deployed in small, measured increments.
Administrators will often start by virtualizing simple or noncritical workloads,
gaining valuable experience with the technology without putting the business
at risk. For example, test, development and QAworkloads are typical targets
for early consolidation work. Once an organization has experience with the
technology, it can systematically virtual-
ize increasingly demanding workloads
with more variable or unpredictable
computing requirements. These might
include workloads such asWeb servers,
terminal servers, SQL and Exchange.
“You need to have flexibility to adapt
your virtualization to workloads,”
Parker said, citing migration tools such
as vMotion or XenMotion.
Ultimately, a virtualized server re-
quires constant resource and performance monitoring to ensure that the sys-
tem is operating within acceptable parameters or to identify bottlenecks that
need correction. But managing workloads and balancing resources can be a dif-
ficult process for administrators. Computing resources are finite, and it’s not al-
ways possible to allocate all of the resources that each workloadmight want.
Administrators need to step back and consider who owns each VM andwhat
SLAs are in place as well as other factors before shifting resources between
workloads. �
9 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
SERVERCONSOLIDATION
With experience,organizations canvirtualize increasinglydemanding workloadswith more variableor unpredictable com-puting requirements.
Capacity Planning
servers are rarely static entities—you don’t just set them and forget them.
Whether the server runs a single operating system and application or shares its
computing resources with a dozen VMs, every administrator has to recognize
that workload demands can fluctuate and grow over time. If the server runs
short of computing resources, the performance and stability of its workloads
can be profoundly affected.
The art of capacity planning allows administrators to ensure that adequate
resources are available in the short term, and it helps them to track changing
workload demands over the long term. Although there is no single process or
protocol for capacity planning, it’s a vital task that every IT professional should
be concerned with.
THE QUICK-AND-DIRTY OF CAPACITY PLANNING
When it comes to servers, capacity planning is the practice of ensuring that the
system provides an adequate amount of computing resources to operate the
workloads that reside there. Computing resources include processor cores or
cycles, memory and I/O for storage and network data traffic. An administrator
relies onmonitoring tools such asMicrosoft’s Assessment and Planning Toolkit
to determine the resources used by each workload and then compares those
total needs to the total resources available on the physical server. Ideally, the
server can provide more than enough resources to service its workloads.
Unfortunately, capacity planning can be a bit more complicated than that. In
a traditional data center where a server operates a single operating system and
application, computing capacity is usually not an issue because servers are
rarely taxed to more than 15% of their total capacity.
For virtualized servers, however, capacity can become a critical issue because
a physical server may host numerous VMs—each using a normal complement
1 0 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
CAPACITYPLANNING
of processor, memory and I/O resources. It’s not uncommon to see a virtualized
server use anywhere from 60% to 90% of its total resource capacity.
But that’s only the tip of the iceberg. Resource needs are rarely static, and a
workload’s needs can fluctuate over time. For example, an accounting applica-
tion may require enormous amounts of processing power to handle accounting
tasks at the end of every month, or a terminal services application may need
more I/O as a company adds users over time. The real challenge of capacity
planning is to watch resource use and see how it trends over time. This allows
an organization to schedule server upgrades, adjust the resources allocated to
VMs, migrate VMs to rebalance workloads across multiple virtualized servers
or take other pre-emptive actions to avoid resource shortages.
And that’s really the underlying point of capacity planning. Resource short-
ages can impair application performance or stability. “The graceful failure
mode is that things just get slower,” said Bob Plankers, technology consultant
and blogger for The Lone Sysadmin. “Sometimes that slowness translates into
time-outs occurring—maybe a database query might time out and then your
appmight crash or experience other unintended consequences.” Capacity plan-
ning uses the data obtained from tools over time, combined with knowledge of
changing business needs, to maintain application performance and availability.
IMPLICATIONS OF CAPACITY PLANNING
Certainly, capacity planning is far more difficult on today’s highly consolidated
servers than it was on traditional nonvirtualized platforms. An administrator
must provide enough computing resources for every workload, enough head-
room—or resources held in reserve—to accommodate increases in resource
demands and still ensure adequate capacity to support workloads temporarily
migrated from other servers. At the same time, the stakes are higher because
resource shortages can potentially affect every VM on the server. “I’ve got 15
physical machines running all of the workloads that used to be on 350 physical
machines,” Plankers said.
1 1 PLANN ING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
CAPACITYPLANNING
Part of this challenge is handled through the use of software monitoring/
planning tools. Microsoft’s Assessment and Planning Toolkit is one popular
option, though virtualization users may opt for tools that are native to the spe-
cific platform they use, such as VMware’s vCenter CapacityIQ. The important
issue here is that tools used on a virtual server should specifically be “virtual-
ization-aware.” Nonvirtualized servers can often rely on operating system-level
tools likeWindows’ PerformanceMonitor. These tools typically provide an in-
stant view of the computing resources being used and—depending on the tool—
may be able to track and chart resource use over time. Based on that data, an
administrator can choose to allocate more available resources to a VM.
When additional computing resources cannot be allocated to a needy VM, an
administrator can opt to migrate the VM to another server with more available
resources. This process of workload balancing is often used to optimize the
number of VMs assigned to any one server by grouping VMswith dissimilar
needs. For example, if you group CPU-intensive VMs on the same server, you
will likely run short of CPU capacity after mounting just a few VMs. But by
watching resource needs andmixing VMswith complementary needs, it can be
possible to fit more VMs on the same physical server and enhance server con-
solidation even further.
MAKING CAPACITY PLANNING WORK
The biggest problemwith capacity planning is the data itself. Experts note that
sound capacity planning requires data gathered and organized over a long pe-
riod in order to be analyzed and evaluated—and it’s simply not a priority for
many shops. “This sort of day-to-day information gathering and analysis … is
tough to do,” said Parker. “Howmany people are truly base-lining stuff over the
course of years or product cycles to give you ameaningful perspective to do ca-
pacity planning?”
And once capacity planning data is available, the decisions to act on that data
are fraught with financial and political pitfalls that can further complicate the
1 2 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
CAPACITYPLANNING
capacity planning process for many administrators, especially in larger organi-
zations where numerous business groups vie for resources. For example,
adding resources to the company’s main database server or customer relation-
ship management servers may be top priorities, but the organization’s file and
print servers or license key servers may
have to wait. It’s important to prioritize
capacity plans and responses based on
business needs and goals. Don’t just al-
locate more resources to any VM or
server that needs it.
Fortunately, not every server is cre-
ated equal. One way to make capacity
planning more efficient is to focus the
tracking and analysis efforts to business-critical servers and then check re-
source use for other servers periodically or as needed. This emphasis also
makes it easier to get the attention and cooperation of other departments or
groups that have an interest in the proper operation of those servers.
Ultimately, capacity planning is an imperfect art, and your capacity planning
process is only as good as the end result. So it’s important to periodically re-
evaluate the effectiveness and success of capacity planning efforts. The more
often servers “break,” the more often capacity planning should be performed.
For example, a stable and static environment with limited growth can probably
make due with capacity reviews every four to six months. But if you routinely
encounter unexpected resource shortages, it’s probably a good idea to adjust the
capacity planning process with more frequent data gathering, more time spent
on analysis and so on. There is no single formula that works for every organiza-
tion, so match the process to your own business needs. �
1 3 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
CAPACITYPLANNING
Capacity planning isan imperfect art, andyour capacity planningprocess is only as goodas the end result.
Benchmarking and Testing
technical specifications are the foundation of any server evaluation.
After all, server hardware must provide enough computing resources to sup-
port the intended workloads—virtual or not. But planning and evaluation can’t
stop with the spec sheet or vendor’s presentation. New hardware investments,
and even well-established systems, demand an objective baseline to ensure ade-
quate performance under real load conditions. Benchmark tools provide this
objectivity. They allow administrators to gauge performance at all points during
a system’s lifecycle andmake critical tactical decisions about how that hard-
ware is deployed.
THE BASICS OF BENCHMARKS
When it comes to defining the behavior of a server, an administrator can use
benchmark or testing tools. In general terms, benchmark results establish a
baseline of performance and utilization. For example, a benchmarkmay tell you
howmany IOPS or CPU cycles are applied to a given workload. “I usually do
that sort of thing when we’re trying to roll out a new application,” said Parker.
Once benchmark data is available, an administrator can alleviate bottlenecks or
tune performance to enhance the workload’s operation.
After an application is evaluated, tuned and considered to be operating nor-
mally, an administrator can use benchmark data as a basis for warnings and
alerts of performance problems. If an unexpected spike in user activity drives
upmemory usage beyond an anticipated threshold, for example, an administra-
tor can receive an alert for action. This can reduce any performance degrada-
tion, forestall user complaints and even prevent the workload from crashing.
By comparison, testing is often regarded as ongoing observation of the
server’s resources or performance over time. By identifying variations in
benchmark or testing results, an IT professional can potentially correlate
1 4 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
BENCHMARKINGAND TESTING
behavioral differences to those variations and then take action to change or cor-
rect them. Periodic testing is also a core part of capacity planning. Similarly, an
administrator may need to evaluate and correct any behavioral differences in
servers or workloads to meet SLA obligations.
It’s important to understand that testing results are often interdependent,
andmaking adjustments to one resource may expose a bottleneck with another
resource. “I eased this storage I/O bottleneck, but now I have a bottleneck in
memory or CPU because now it’s not standing around waiting for storage I/O
all the time,” Parker said, noting that changes to the application’s resource use
must be documented and new alerts set to reflect any new resource settings.
MAKING THE MOST OF BENCHMARKS
Just choosing the right tool can be a challenge itself—there are countless bench-
mark tools available. Citrix EdgeSight for Load Testing, Iometer for storage per-
formance and IOzone for Linux environments are just a few of the many
potential candidates for consideration. Other benchmarks specialize in specific
tasks, such as a UDP simulator to benchmark UDP network traffic, whileWeb
servers may rely on tools designed to simulate HTTP, AJP, JDBC and different
load types. Regardless of your specific needs, experts urge care in selecting the
right benchmark tool for the job.
One of the biggest issues is the influence of vendor-specific benchmarks, and
experts warn to verify the objectivity of benchmark tools before making a deci-
sion. “Occasionally people have suggested that maybe the information about
performance in places might be slightly biased in favor of what the vendor
would like it to be,” Parker said. “For example, it says there isn’t a problem here,
yet when I’m on these virtual servers it does seem like things are a little slower
than they should be.”
When server virtualization is deployed in the data center, be sure to select a
benchmark or monitoring tool that is virtualization-aware. This means the tool
is designed to “see” the server’s virtualization layer and understand the way
1 5 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
BENCHMARKINGAND TESTING
that computing resources are allocated and tracked across VMs. A tool that is
not virtualization-aware may produce skewed or inaccurate results that might
lead to problems with server resources later on.
Also consider the way that benchmarks andmonitoring tools are installed
and used. Ideally, a tool is relatively lightweight—requiring few computing
resources to accomplish its tasks—and produces results that are reasonably
accurate when compared to the results of
other similar tools. It avoids tools that rely
on agents or any deep connection to your
infrastructure. Tools that violate these
basic principles should be approached
cautiously by IT administrators.
When configuring benchmark and
monitoring tools, pay attention to the res-
olution of any collected data. For example,
if you’re looking for peak resource usage
over the course of weeks andmonths—such as end-of-month accounting sys-
tem demands—gathering data every fewmilliseconds is probably not useful.
Conversely, trying to spot an unexpected resource peak that is causing alarms
each night a batch job runs would demand far more frequent data points to
analyze.
And finally, gathering benchmark and performance data is useless unless
youmake the time to sit and actually analyze the data—that’s a glaring over-
sight in so many hectic IT departments. “Gathering that data is one of the few
things that really lets me domy job well,” Plankers said. �
1 6 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
BENCHMARKINGAND TESTING
Gathering bench-mark and perform-ance data is uselessunless you make thetime to analyze thedata.
Stephen J. Bigelow, a seniortechnology editor in theData Center and Virtualiza-tion Media Group at Tech-Target, has more than 20years of technical writing ex-perience in the PC/technol-ogy industry. He holds abachelor of science in electri-cal engineering, along withCompTIA A+, Network+,Security+ and Server+ certi-fications, and has writtenhundreds of articles andmore than 15 feature bookson computer troubleshoot-ing, including Bigelow’s PCHardware Desk Referenceand Bigelow’s PCHard-ware Annoyances.
1 7 PLANNING A SERVER REFRE SH
SERVER HARDWAREROADMAPS
SERVER CONSOLIDATION
CAPACITY PLANNING
BENCHMARKINGAND TESTING
Cathleen GagneEditorial Director
Jeannette BeltranAssociate Managing [email protected]
Marty MooreAssociate Managing [email protected]
Linda KouryDirector of Online [email protected]
Marc LaPlantePublisher
TechTarget275 Grove Street, Newton, MA 02466
www.techtarget.com
© 2011 TechTarget Inc. No part of this publica-tion may be transmitted or reproduced in anyform or by any means without written permis-sion from the publisher. For permissions orreprint information, contact Renee Cormier,Director of Product Management, Data CenterMedia, TechTarget ([email protected]).
ABOUT THEAUTHOR
RESOURCES FROM OUR SPONSOR
• The cost of delaying server refresh cycles
• Sizing Servers: How to Create Your Ideal Infrastructure
• Dell PowerEdge Energy Smart Servers -- The Smart Choice for Optimizing your Data Center
About Dell, Inc. and Intel:Dell and Intel are strategic partners in delivering innovative hardware solutions to solve yourmost challenging IT problems. Together we are delivering virtualization optimized solutions tomaximize the benefits of any virtualization project.