Virtualisation performance management - Myths and realities

Performance management has become the key element in maximising any virtualisation initiative, but many myths have emerged that, left unchecked, can stall an entire virtualisation initiative. Here are the biggest myths and the corresponding realities to enable IT to chart an informed path forward and successfully scale deployments.

1. VI and desktop administrators only need to worry about server metrics.

Reality: Performance management tools must be able to analyze information across all IT silos - from the end user, to the application, to the network, to servers, VMs and storage - covering both the physical and virtual metrics. Without this data, administrators cannot accurately pinpoint and troubleshoot potential performance issues, such as when there's a clash of resources across the WAN or a storage latency issue.

2. It is impossible to fully capture the end user experience as it happens.

Reality: Performance management solutions must capture the end user experience in order to ensure end user satisfaction, identified as the No. 1 factor in determining the success of any VDI implementation. If users can trigger a DVR recording of their activity when problems occur, administrators can capture the real-time load on the infrastructure, such as CPU, memory, storage and the interactions of the desktop. A recording like this gives users a simple and proactive way to immediately communicate with IT as soon as they encounter performance problems - instead of trying to explain what went wrong after the fact.

3. Performance management doesn't have to be live.

Reality: To ensure performance of a virtualised infrastructure, you must deal with information associated with a huge number of objects in a live and continuous fashion. Without live, interactionally focused performance management you only have partial picture of what's happening in your infrastructure. Second-by-second insight allows IT to see what set of interactions caused a particular performance shift; this 'liveness' allows for triage versus post-mortem analysis.

4. Capacity planning and performance monitoring deliver the same functionality.

Reality: Capacity planning and performance management provide complementary capabilities at opposite ends of the time scale. Capacity planning is useful to identify long-term, slow-moving trends that may affect production systems in the future. It can also be used in scenario-planning exercises to determine things such as the number of additional desktops that can be added to a VDI cluster. Performance management, on the other hand, is the only way to catch fast-moving contention events that occur in mere seconds or minutes.

Operationally, performance management is used every day to inspect the infrastructure as it dynamically shifts and flexes to accommodate user and application load. In production networks, performance management covers the immediate day-to-day requirements of understanding, troubleshooting and optimising virtual data centers; capacity planning provides a long-term, future-facing view of slower moving events.

5. Virtualisation is just another silo.

Reality: A virtualised environment is all about shared resources, as opposed to a separate and distinct silo. In virtualization, the silo walls have come down, and a holistic approach toward performance management is needed to manage these shared resources. IT must fully understand how virtualised applications impact other workloads to get the most value from their virtualisation initiatives.

6. Agent-based software is necessary.

Reality: Agent-based software limits the scalability of virtualised environments - increasing bottlenecks and duplication. Having a guest agent in every VM severely restricts IT's ability to scale their virtualised infrastructures - and limits the extensibility that is required in order to accept data feeds from multiple silos.

7. Separate tools are OK in a virtualised infrastructure.

Reality: Understanding the causality of performance issues is critical in virtualisation. IT needs to have understanding about the causality of performance issues, and to do so requires integrated insight in a single system - for example, insight into all objects' interactions with other objects that, in fact, may be causing the problem. Virtualisation is a dynamic infrastructure, and IT needs to monitor dynamic shifts as they occur to see correlation between objects and interactions.

8. Virtualising line of business applications is too risky.

Reality: Companies are realising that virtualising their LOB applications is absolutely necessary for them to achieve the full benefits of virtualisation. With live and continuous, cross-silo insights spanning the physical and virtual, IT now has the understanding needed to ensure the performance and availability of their most critical applications - stopping the stall and maximising the value of initiatives.

9. ROI for performance management is impossible to determine in virtualisation.

Reality: Effective performance management gives IT a way to maximise the full ROI of their virtualisation initiatives by reducing time to problem resolution and extending their virtual footprint to include LOB applications. Both enhance virtualisation's value to organisations and give IT the ability to more easily communicate the long-term value of the technology beyond capex savings in terms of hardware.

10. Large IT management vendors will lead virtualisation management.

Reality: Managing virtualised environments differs in almost every way imaginable from managing a physical legacy environment. Third-party management vendors - over enterprise management vendors - are ahead of the game with the functionality - and advanced interactional analytics capabilities - designed specifically to manage dynamic virtualised environments, as well as growing hybrid infrastructures with multiple hypervisors.

Nathanael Iversen is director of technical marketing Xangati. This vendor-written tech primer has been edited by to eliminate product promotion, but readers should note it will likely favour the submitter's approach.