Vic Nyman

Business Trends Quarterly -- Virtual IT

This article originally appeared in Business Trends Quarterly, May 2009.

 

 

To generate and achieve full ROI from virtualization efforts, organizations are now virtualizing servers that support business-critical applications. While many tools help manage virtual servers, they provide little real-time, actionable data on how the applications themselves perform and how these virtualized applications interact with one another and the infrastructure in which they reside.

Application owners and those responsible for keeping critical applications performing well are concerned and pushing back on projects to virtualize the hardware on which their applications reside. Business owners and application support personnel have concerns over the potential impact on performance and the manageability of their applications. They need assurance that their applications will perform as well – if not better — on a virtual machine (VM) as it did on a dedicated server. To ensure that their complex business applications are functioning properly in a mixed virtual/physical server environment, IT managers must take an application-centric approach to managing and optimizing the performance of critical applications.

Manage the application, not the machine.

As virtual application performance is abstracted from the physical server hosting the VM, support organizations must quickly determine and isolate where an application is bogging down. However, many organizations do not have the visibility necessary to monitor what or how their applications are doing on their VMs. They can measure and report symptoms, but cannot diagnose the cause. Many management tools look only at the VM, the server, or the network, taking a silo approach to problem identification. These techniques fail to show where the application goes, the shape of its infrastructure (virtual, physical, or both) and only provide a fraction of the performance details required for genuine problem solving. Since virtualization breaks the one-to-one relationship of server-to-application, organizations can no longer solely rely on machine performance indicators to determine the health of their applications.

With an application-centric approach and the proper tools to see program interdependencies, application owners and IT support teams can keep complex applications performing well. This approach provides performance details of service-level processes, depicting the following: application response times, application details of resources used, bytes sent and received, and processes maintained, dropped or stalled, as well as detailed metrics on machine performance.

Application relationships must be mapped and monitored across servers and operating systems throughout the enterprise. Only by interacting at the application service level, where full details into the performance and behavior of the application are available, can it be seen where the application is impacted, and what is impacting its performance. Once that is understood, application support staff can dive into the server stack to determine if there are bad connections, an overloaded VM, server-hosting conflict, or any number of server-related issues. Here’s an example where bad connection logic highlighted issues with virtualization and application visibility:

Case of the Lost Connection

A company virtualized much of their data center. Suddenly, a highly used SharePoint application became effectively non-responsive. After a few days of work, and numerous finger-pointing meetings between VM Administrators and application owners, the applications support team realized that a server the application was dependent upon had been converted to a VM; yet their application was still making calls to the old address.

What should they have done differently? With an application-centric approach, which would have mapped shifting application relationships as they migrated to virtualization, they could have quickly followed the application, isolated the change and the lost process connection, and saved themselves days and tens of thousands of dollars in downtime and diagnostics.

Understanding application requirements in P2V projects

Virtualization can add a processing strain by offloading network I/O, which goes beyond simply stacking too many VMs on top of a host server. Virtualization causes network I/O – and often storage I/O – to be handled multiple times by the same CPU complex. This generates new CPU overhead directly associated with I/O functions.

As the transactions and dependencies involved grow (such as those associated with n-tier business applications and SOA deployments, for example) P2Ved applications do not scale as simply as one might expect. Applications
and infrastructure teams need to do their homework on application I/0 so they can prepare for increased CPU utilization in advance of a P2V conversion. To accomplish this, they need tools that can monitor these issues in real-time as their VM environments change.

P2V Redefined = Prior to Virtualizing

A major telecommunications provider wanted to identify and understand what applications and services were resident on servers identified for consolidation, prior to making any changes. By seeing the existing interdependencies and understanding what the application was touching and required they could have avoided any applications becoming “P2V Collateral Damage.”

By employing specialized application discovery and mapping tools, users follow their applications across the infrastructure identifying dependencies and connections — prior to the P2V process.

This approach ensured that their critical business applications remained online throughout the conversion process. They verified performance metrics before and after the P2V, certifying that their applications were ready for production use. Once converted, they measured the application performance within these consolidated environments, ensuring that the business-critical applications continue to perform well, well after P2V conversion.

The application-centric approach becomes an imperative as virtual server environments grow more complex. Added processes need to be seen, VM changes need to be monitored, and detailed performance data needs to be available for every monitored connection so application visibility is maintained, regardless of the latest data center change.

 

The cost of application performance issues

Application downtime or poor performance hurts – virtualized or not. However, the lack of application visibility in virtualized environments has caused some organizations to halt their P2V plans for critical applications altogether. In this example, an application-centric management approach helped revive a stalled P2V initiative.

Over-provisioning Solves Nothing

To address application performance issues and minimize downtime, IT staff at a leading software provider were buying more capacity and provisioning VMs with more resources to keep things running. When performance issues become evident, they would provision another VM or add resources to existing VMs, over-provisioning them to the point of wasting precious resources and defeating the purpose of virtualization in the first place.

What is the solution? By using new tools that offer visibility into apps as they reside on virtual servers, support teams can determine what performance behavior is typical for the application. They can understand what resources are required to keep the performance running within normal ranges, right-sizing the application with appropriate resources.

When problems do arise, they can follow the application across the infrastructure, drilling into the server stack for targeted investigations and diagnoses. They are able to quickly isolate problems and eliminate the need for additional resources, ultimately meeting user requirements while optimizing their resource pool.

New tools address emerging challenges

Today, a new generation of Application Service Management (ASM) tools are capable of providing real-time, actionable application data — a first for the virtualized data center environment. This gives application owners the confidence to virtualize complex applications, knowing they will perform as expected.

These solutions use intelligent data collectors (a small service that runs in the operating systems of a few servers in the environment) to passively collect data by monitoring the application request layers. The collector sends detailed statistics on the operating systems by using WMI, when necessary, and ESX Server, by connecting to VMware VirtualCenter management APIs. This data helps application administrators understand application performance and track its movement in order to quickly detect potential bottlenecks and failures.

 

Application Service Management (ASM)

ASM is a better approach to managing applications in today’s dynamic data centers. An example of an ASM solution is BlueStripe’s FactFinder solution. FactFinder provides application support and infrastructure that delivers visibility to manage the performance and availability of applications deployed in virtualized data centers. ASM solutions include three critical elements:

• Application discovery & mapping - Discovery and mapping of application connections, processes, and interdependencies, allowing for visibility into both physical and virtual environments.
• Service-level health & performance measurement - Performance intelligence detailing response times between nodes, all connections, and application access times, depicting the relative health of complex applications.
• Triage of application & server problems - Performance indicators highlight areas of concern, showing where to focus efforts to solve problems and optimize the IT environment. Following the application response times across the application infrastructure to identify where performance degrades, then down the server stack, enables quick triage of complex problems.

Since ASM stresses a business approach, application support and infrastructure owners can align service-level objectives around overarching business priorities. This application-centric approach helps clarify goals across organizations and improves responsiveness.

As a result, application and infrastructure support can focus on the availability, performance and mutually defined parameters of business processes, as well as specify the processes through service-level agreements. Essentially, ASM gives IT teams a common language to share objectives, and provides visibility into actual application performance and availability, allowing them to control and improve service delivery to the customer. BTQ