Taking the guesswork out of capacity allocation

Getting IT capacity planning right need not be a question of taking your best guess, if organisations use the proper mix of APM and capacity management, says Sanjeet Padhy, Practice Director, CA MENA

Tags: CA Technologies (www.ca.com/us/default.aspx)
  • E-Mail
Taking the guesswork out of capacity allocation Padhy: The tools are available to accurately predict capacity requirements without resorting to speculation.
By  Sanjeet Padhy Published  November 21, 2013

How many times have you sat in a restaurant and said, “The service is a bit slow in here”? Chances are you walked out, decided not to eat there again, or advised friends not to go near the place.

Blame the people that run the restaurant: they probably didn’t consider the number of customers that would turn up, and didn’t staff up accordingly. But let’s not be too harsh. You could walk into the restaurant on another day and it might be milling with staff staring at empty tables. Maybe you caught them on a bad night.

Your IT department can feel like that restaurant sometimes. Your customers — in this case the business executives, external customers, internal employees and others — all want their applications to work fast and flawlessly. And another customer — the CFO — also doesn’t want to pick up an unnecessarily big bill at the end.

It all comes down to capacity planning: being able to reliably predict how much IT capacity is needed based on past application performance so you can organise your infrastructure accordingly. Get it right and you deliver an exceptional end-user experience, reduce cost and minimise risk. Get it wrong and you risk a loss of business through application downtime, high capital expenditure (CapEx) costs owing to an under-utilised infrastructure and uncertainty over how effective your infrastructure is at delivering service levels.

Planning capacity is a big challenge. Your budgets are as flat as the atmosphere in that under-staffed restaurant, while the demand for IT services has sky-rocketed, driven by the march of mobile services, cloud and the consumerisation of information technology. Moreover, complexity is everywhere: a blend of physical, virtual, cloud and mainframe systems all need to be optimised to deliver business-critical applications.

First solution to the problem? Overprovision. Do what the restaurant manager would do, and build up your resources (in this case, your infrastructure) to meet the forecast increase in demand for IT services. You probably won’t be in business for long if you do though. If the demand doesn’t materialise the way you think it will, you are left with idle servers. Over-provisioned data centres not only call for higher CapEx, they consume more operational costs, like maintenance, upgrades, power, and licensing. And at the back of your mind is the knowledge that the average utilisation of a virtualised server can be as low as 20% range.

Application performance management (APM) has most of the answers. APM captures your transaction performance data from problem sources like applications, end-users and the infrastructure, and uses integrated end-user experience information to help prioritise problem resolution and manage service level agreements. However, you still need a way to cost-effectively address the capacity issue without increasing risk to the business.

Unified predictive capacity planning

The real answer lies in blending the power of APM with the control of capacity management. By leveraging both APM and capacity management in a unified predictive capacity planning solution like the one offered by CA Technologies, you can more reliably forecast future capacity needs. This allows you to mitigate risks, help ensure quality of service and right-size your application delivery environment while optimising costs.

It works in two ways. First, when a problem occurs, APM alerts IT operations to an incident — like a server running at more than 80% utilisation. You recognise the need to take action and reach into the server workload data. You then use the capacity management component to run ‘what if?’ scenarios for the affected server and entire application delivery chain, solving the system problem which caused the APM alert.

Take the real world case of a multinational food and beverage company. Their existing testing process failed to uncover bottlenecks in the company’s large-scale SAP environment. Testing was also a very expensive process and the company was under pressure to cut costs and improve service delivery. This food and beverage company deployed a predictive reliability solution to deliver an early warning system during the design, test and production phases. This meant the development teams could discuss the data, redesign, and avoid problems that could have occurred in production. As a result, the company saved significant time and money by catching design problems early in the lifecycle.

The second scenario for predictive capacity planning involves right-sizing your environment for future growth. Using APM performance data from your production environment, the solution enables you to conduct scenario analyses simulating different load patterns. This allows you to optimise your production infrastructure with the right system configurations based upon the planned workload.

That’s how a leading financial services organisation uses predictive capacity planning. The company was challenged to adapt to changes in consumer expectations — always there, always on, always with me — and also needed to lower the cost to build and operate applications. Model-based performance testing provided the confidence to know what’s going to happen, for any given change to the finance firm’s application environments. Costs went down, performance went up, and confidence in the supportive infrastructure soared.

Guessing how capacity is allocated and consumed is never going to be the answer. Predictive capacity planning takes lets you reliably right-size your infrastructure based on real performance trends. No conjecture, speculation, or guesstimates.

Add a Comment

Your display name This field is mandatory

Your e-mail address This field is mandatory (Your e-mail address won't be published)

Security code