A strategy to leverage virtualization to provide the most cost-effective infrastructure, while successfully deploying new applications, is top of mind for the CxOs in many corporations. Properly done, virtualization cuts costs while facilitating new revenue streams. Improperly done, it leads to devastating contention storms that can destroy the reputation of your brand and customer loyalty in the blink of an eye. This is usually the result of applications running on a shared infrastructure that can adversely affect each other through unintended and unpredictable resource contention storms.
Before signing off on your plans to migrate critical line-of-business applications and your desktop infrastructure to a virtual infrastructure, be sure you deliver the key four must-haves with your performance management solution to instantly identify and resolve application performance issues caused by the complex infrastructure they run on. They are 1) cross-silo, 2) live and continuous, 3) scalable and 4) interaction-focused.
Sharing to achieve leverage leaves virtualized infrastructures with little or no separation between silos. The days of distinct, individually managed silos are fast becoming a relic of the past. With a virtualized environment, the center of gravity is not found in just one specific silo. Rather, there is continuous and dynamic interplay occurring within and across multiple silos. For example, multiple virtual machines (VMs) representing an application can dynamically interact on a single server or storage device.
To adequately address this large-scale dynamism across silos, any performance management solution must be comprehensively cross-silo. This unique vantage point can correlate cross-silo information to understand and act on unified information as to why, for instance, storage backups are taking longer than expected or VDI end-user experience is being impacted or how memory is being utilized. Only by viewing and correlating cross-silo information – compiled from all silos: network, storage, servers, VMs, applications, etc.— will you have concrete and actionable context for root cause analysis and recommended remediation.
Live and Continuous
In a virtual environment, understanding what is happening at any precise moment in live and continuous manner is critical in order to uncover and fully understand the causality of toxic behavior and interactions between objects in the virtual infrastructure. Without second-by-second insights, the live and continuous transient problems are missed completely. Reports that are averaged over minutes or hours, or as long as a day, miss these vital insights. Successful performance management needs to see dynamic shifts as they occur and how they change over time to identify and correctly resolve many problems.
The challenge is to accomplish real-time at scale. The amount of data in a virtualized infrastructure is voluminous, and the amount of time you have to process that information is minuscule – no more than a few microseconds for each interaction. Solutions need to scale to millions of objects and interactions in a virtualized environment, rendered on a per-second basis and must analyze the data in the time span of a single microsecond.
Virtual environments grow like there’s no tomorrow, with objects and interactions scaling to 10s and 100s of thousands! The requirement to be live and continuous is closely tied to the requirement to be highly scalable. You not only need to know what’s happening second-by-second, you also need to be able to do so for a very large number of objects. Look for a memory-driven architecture that allows administrators to see dynamic object interactions scaling to 100s of thousands.
Focused on Interactions
It is insufficient to understand only the consumption of various resources by applications at any given time. Effective performance management requires an understanding of how the actions of one object affect another, within and across applications. A problem with one object will impact other objects that interact with it, potentially causing the problem. Analytics need to detect and show the impact of interactions, such as a CPU spike related to a certain interaction between two VMs, for example. Only by understanding the effect of the combination of specific interactions can a performance management solution truly help to identify the particular root cause of a contention issue.
The ultimate key to successful execution of applications is the ability to manage in real-time so that the IT department can perform triage, not post-mortems, when the infrastructure-caused application problems arise. By automatically unearthing performance problems as — or even before — their adverse effects are felt, administrators solve and/or contain these problems before their effects ripple through the infrastructure and become harder and more expensive to resolve.
In an increasingly complex and diverse virtualized environment, application performance is affected by the dynamic contentions in the shared infrastructure on which they run. Thus your performance management solution must know, in a live and continuous manner, about each and every object in the infrastructure across silo, how they behave and interact with each other to a scale of hundreds of thousands of objects, when and why the interactions result in resource contentions, and how to resolve them by infrastructural reprovisioning or expansion. These actionable insights can potentially save your enterprise significant and unnecessary costs and help obviate lost revenue.