Dec 28 ’11
Moving Capacity Planning Forward With zEnterprise and Private Clouds
Capacity planning today is anything but “business as usual.” Forward-thinking organizations are already redefining capacity planning as “capacity management” because performing capacity work effectively in an environment that will feature on-demand provisioning and private cloud services involves more than just looking at MIPS consumption for annual budget planning.
The movement to private cloud compels some rethinking about how IT is organized. IT staff and resources must now support business services and accompanying Service Level Agreements (SLAs). If IT is to rally staff and resources around business service support, departmental functions historically isolated must come together under each service, whether they originate from CICS, DB2, WebSphere MQ, network management, mainframe, or distributed platforms. Staff members responsible for these functions must collaborate to plan for and execute delivery of business applications and services across many platforms. Their success requires capacity planning and measurement tools that provide both broad and granular visibility.
Where Capacity Planning Fits
Capacity planning is a component of capacity management, which also encompasses resource optimization, tuning, and performance reporting. More companies realize capacity management offers much greater value when actively used in ways that transcend capacity planning.
A strong business case supports this philosophical change. SLAs, which often accompany the move to private cloud implementations and a business service delivery approach in IT, are forcing IT to focus on how its resources are delivered. Part of this process is ensuring there’s adequate service level resource capacity. Where capacity is insufficient, IT must address it. Resource capacity planning also emerges when unplanned upgrades are necessary, as in the case of a business acquisition or introduction of a new line of business.
Because of these and other factors, more companies are re-evaluating the role capacity planning plays and coordinating the effort with the adoption of Information Technology Infrastructure Library (ITIL) best practices.
The Impact of zEnterprise
The new zEnterprise cross-platform management capabilities enabled by the IBM zEnterprise Unified Resource Manager (zManager) with its firmware-level integration present capabilities that are further enhanced with end-to-end application management software. The introduction of these capabilities as organizations adopt private cloud implementations has raised awareness that capacity management should occur at the business service level, regardless of the platforms a given application crosses, and organizations should take a broad view of the process. This is a step forward from traditional approaches where capacity planning was limited to individual subsystems such as DB2, CICS or WebSphere MQ, or to individual platforms such as mainframe and distributed systems. The more holistic approach now emerging considers the impact on overall infrastructure service delivery to the business. zEnterprise is a natural enabler to this approach, which facilitates the service orientation private cloud computing demands.
Systems management software for zEnterprise includes unified toolsets that let all staff experts, whether they command DB2, CICS or some other area of IT infrastructure, use the same information and customize it on individual dashboards. This makes it easier for IT departments to collaborate on service delivery and manage capacity. A common toolset and common data eliminate the potential for confusion that can occur when disparate tools and results measurement methodologies are used.
As IT breaks down silos and migrates to enterprisewide capacity management that’s done at the business level, capacity planning and management can effectively be deployed to model future business scenarios, too. It could work like this:
If the claims transactions processing service level for an insurance company is two seconds, IT’s capacity management system should be able to tell this to IT and business managers. Capacity management should also indicate if the company is getting ready to cross a threshold where a service level is threatened. If the company is bumping up against the service level, capacity management tools should be able to tell IT whether the problem can be solved with performance tuning, or if it requires an upgrade. In all cases, the goal is to predicate technology decisions in capacity management on business decisions. This means capacity planning and management must be linked to business planning.
Capacity planning and management should also be able to be used for futuristic scenarios. If, for example, the business projects claims will increase by 25 to 50 percent over the next 18 months, capacity planning models should be able to take that information and project what the resource pool will look like under those increased transaction scenarios, and what types of actions or investments are necessary.
In all these business examples, various disciplines in IT will likely end up together in the same meeting room with the business. IT will need to further embrace its ongoing shift away from isolated problem solving. A smart capacity planner will reach out to all parties as part of the planning process and discuss increases in transaction volumes and Key Performance Indicators (KPIs). He or she will translate business strategies directly into IT strategies without encumbering the dialogue with lots of technical jargon.
As sites implement private cloud deployments, capacity management across platforms for the delivery of singular business services will become even more important. So, too, will be defining the role zEnterprise will play in the cloud. That role is still largely undefined in most organizations, even though there’s a common commitment to plugging zEnterprise into the cloud because of its ability to manage and plan for capacity across mainframe, UNIX AIX and x86 resources under “one roof” with a single set of tools.
By using zEnterprise as part of their private cloud solutions, sites can take advantage of its extensive virtualization capabilities and sophisticated management practices. On the distributed side of IT, there are some larger VMware boxes that virtualize, but their capacity management capabilities remain evolutionary. What they’re aiming for is a Parallel Sysplex approach like the mainframe’s—where they can dynamically manage workloads across a variety of environments.
Sites are still acclimating to the implications of internal cloud computing, business service delivery, and being measured against SLAs in the same way as commercial vendors. Understandably, they haven’t all made the transition to capacity planning that’s integrated across departmental functions and all of the disciplines that comprise capacity management.
A common mistake sites continue to make is basing their capacity planning solely on MIPS and not on business goals. Because sites don’t do a lot of work linking IT resource capacity to the business, CPU sizing typically occurs when the site intends to make a move, such as a hardware upgrade. In this context, capacity planning is a highly focused function.
The alternative is to move to a business service delivery environment where actual transaction volumes are tracked against capacity to ascertain performance. This should include trending information that maps resources to business volume increases. Capacity planning and management that occur in this manner helps IT take a broader look at how business system resources are functioning and what the capacity projections will be in the future. It can also help determine the impact on system resources if a new line of business debuts; for example, if you suddenly ship more packages or execute more trades, this affects many subsystems (WebSphere MQ, Batch, CICS, DB2), so you need to know upfront whether you will be building a breadbox, a garage or a building, and you need to understand the implications on resource capacity at a granular level.
A second area sites underestimate in capacity planning and management is application changes. The applications staff doesn’t usually say, “We changed all this,” but you might see more draw on resources with a new application release. Sites often neglect analytics modeling for capacity planning that examines I/O and CPU performance in applications. Instead, there’s a tendency to just put applications in production environments and see what happens. In sharp contrast to this, one site lowered CPU utilization 46 percent by tuning applications—and postponed an upgrade.
The following best practices can help IT attain and maintain the capacity management capabilities needed to support private cloud implementations, SLAs, and the ongoing evolution of IT infrastructure:
- Align capacity planning with the business. Involve business stakeholders and include business KPIs in discussions.
- Focus on business service level commitments, not MIPS alone.
- Use capacity sensitivity analytics to plan for unplanned events or those with a significant impact (e.g., the sudden promotion of a new product that generates high transaction volumes or a corporate acquisition that does the same).
- Continuously track and trend actual vs. planned capacity, and when possible, use capacity planning capabilities to effect continuous improvement of systems through system and application tuning opportunities.
- Move more structure around ITIL best practices.
- Consider using more zEnterprise specialty processors such as System z Integrated Information Processors (zIIPs) and System z Application Assist Processors (zAAPs). Some organizations worry about shifting workloads to these processors due to performance concerns. However, capacity planning analytics and modeling can help determine the resources appropriate for these processors, which can provide significant cost savings.
- Pursue the integration of specialty technical disciplines and platforms within IT and collaboration between business and IT.
Capacity planning and management are anything but stationary in an IT environment that’s moving toward private clouds and business services. What matters now is that IT takes steps to assure capacity planning keeps up with IT infrastructure evolution. It can do so by adopting certain best practices that facilitate a nimble, holistic approach.