Performance and Capacity Assurance in the Virtual World

   
Lisa Crewe is director of product marketing at Akorri.

From an operational perspective, virtualization introduces some new challenges and makes others worse. For example, capacity management is a mystery.

For that reason, it has been described as a “dark art.” According to Forrester, capacity management is the top operational concern among IT professionals using virtualization.

Source: Forrester, Dec 2009

Capacity Impacts Performance

By abstracting the virtual servers from the physical hardware, it becomes much harder to determine the actual resource requirements. This introduces confusion and risk around the performance and stability of workloads intended for virtual servers.

As a result, many workloads remain on physical servers because the risk is perceived as too high. However, with the right management solution, you can manage capacity and assure performance of business-critical (Tier 1) virtualized applications and save time and money too.

Is the Hypervisor Manager the Right Solution?

Most companies that have gotten past that first stage of virtualization realize that hypervisor managers, like VMware vCenter Server, alone don’t assure performance of business-critical applications because of the dynamic and shared nature of the virtualization environment and the mix of virtual and physical resources in the infrastructure.

To effectively assure performance, the management solution must comprehensively manage the other components of the overall virtualized service, which are supporting the virtual server, including physical systems and storage. These resources, previously living very separately, are now increasingly interdependent as a result of virtualization.

Top Three Virtualization Performance and Capacity Challenges

These are the main reasons why you need a management solution over and above what you get with the hypervisor manager:

1. In a virtual environment, it’s impossible to assure application performance based on how applications use resources.
2. A virtual environment grows and changes quicker than a physical environment.
3. Most environments are mixed (virtual and physical).

What’s Required?

In order for a performance and capacity management solution to be effective, it must meet the following high level requirements:

• Map the topology that it is monitoring dynamically from end to end.
• Leverage the resource utilization data available in the vCenter APIs.
• Provide a rich set of capacity planning and management functionality.
• Provide infrastructure response time for the relevant elements of the entire system.
• Deliver infrastructure service management leveraging both the resource utilization data and the infrastructure response time data to ensure that the infrastructure is providing the required level of response time to business critical applications.

Why Infrastructure Response Time Is Critical

Infrastructure Response Time is defined as the time it takes for the system (virtual server, physical host, storage) to perform work submitted by an application and thus it directly shows the system’s infrastructure responsiveness.

Infrastructure Response Time can be used to quickly show if poor end-user performance is due to infrastructure issues. If Infrastructure Response Time is poor, the end user’s experience will likely be poor, but if Infrastructure Response Time is good, when the end user’s experience is poor, there is likely a problem with a non-infrastructure component of performance. Tracking Infrastructure Response Time is a great way for IT infrastructure owners to prove that they are providing good infrastructure service to application owners, and can be used as a base measurement for service level agreements (SLAs).

Understanding VMware’s Management Offerings

Focusing on the performance and capacity capabilities in VMware’s vCenter Server, CapacityIQ and AppSpeed, let’s see how they’re different from what we just outlined as key requirements above.

vCenter Server: Daily Management; Basic Resource Monitoring
vCenter Server is a required component of VMware’s hypervisor if you want to take advantage of the underlying value of the hypervisor. It is effectively the window into the virtualized data center. That’s the tool that enables all of the virtual machine management and server management and storage provisioning tasks to the virtual machines, and all of the other day to day stuff required to control and maintain the environment. It’s also an extensible platform for VMware partners to plug into. VMware makes a rich set of resource utilization data available via the vCenter APIs. Many vendors take advantage of this data and store it, trend it, analyze it, report on it, and alert on it. However, this data tells you only how the resources in the vSphere environment are being used, and do not provide information on either infrastructure performance or applications performance.

vCenter CapacityIQ: Resource-based Capacity Planning and Management
vCenter Capacity IQ, as the name suggests, performs capacity planning on virtual infrastructures, applying continuous what-if analysis to figure out the best arrangement for virtual machines in different scenarios. It offers reporting and recommendations. CapacityIQ is made of two components: a vCenter plug-in and a virtual appliance that collects data about the virtual infrastructure in a dedicated database.

So, how is it different? The ideal performance and capacity management solution includes I/O workloads as part of its analysis into bottlenecks and optimization.  CapacityIQ only looks at CPU and memory. CapacityIQ can help with the easy low-hanging fruit, such as idle VMs, but cannot assist in optimizing VM layout or density. CapacityIQ is limited in analyzing simple metrics from vCenter without any correlation to other elements in the infrastructure. True optimization is based on utilization and performance; CapacityIQ only looks at simple utilization. In addition, CapacityIQ predicts when utilization thresholds will be reached, but does not take into consideration future growth requirements. It also does not provide information on storage or physical servers.

vCenter AppSpeed: Application Performance Management
VMware AppSpeed is actually an application performance management solution rather than an Infrastructure Performance Management solution. It is intended for use by the App Teams and is a virtual only solution. In other words, it has no ability to monitor application systems that span a virtual and a physical infrastructure.

Assuring Performance of Business-Critical Applications
True performance and capacity assurance in a virtual world, requires insight and analysis into both utilization and performance for virtual and physical elements in the infrastructure. Infrastructure Performance Management, which includes the ability to leverage data from the hypervisor and augment it with unique analysis to create metrics that measure all of the underlying resources in the infrastructure is required in order for IT Operations to successfully virtualize Tier 1 applications. The key metric for Infrastructure Performance Management is Infrastructure Response Time and should be present in any effective solution.

Without the Infrastructure Performance Management solution, the IT Operations teams supporting the infrastructure will continue to practice the “dark art” of guesswork rather than confidently assuring performance and making smart decisions on capacity that deliver value to the business.

Lisa Crewe is director of product marketing at Akorri. She has more than 15 years of experience in product marketing, channel marketing and marketing communications for high-tech companies. Before joining Akorri, Lisa held director of marketing positions at enterprise networking company Converged Access and telecommunications equipment provider Reef Point Systems. Prior to that, she was senior manager, product marketing at Lucent Technologies where she was responsible for worldwide marketing of wide area network telecommunications equipment.