High Volume Risk: How to avoid desktop virtualization outages during peak load
High load testing represent a challenge to both product development teams and to IT as highlighted by last week’s outages at the NYSE, United Airlines and the WSJ. Industry best practice is to load test all software updates, and infrastructure at peak volumes to understand behavior and be certain of scalability. But in reality, this investment is typically weighed against the likelihood of an outage and is not fully committed to, or the proper time cannot be allocated to thorough, continuous, full load testing. The NYSE, United Airlines, and the WSJ are all large enterprises with significant resources to commit; so why did these public outages occur?
Product development teams and IT do their best with the resources they have to successfully test and deploy. Many methods are in use today, and these methods encompass the full gamut of the testing life cycle - smoke, functional, systems integration, regression, user acceptance, etc. But today’s systems are complex with an inordinate number of moving parts; parts which may be shared and whose behavior under peak load needs to be understood prior to production release. This is true for any production system, including those that are virtualized. Under duress, updates to some of these moving parts may be made outside of standard operating procedures, introducing risk to systems stability, especially under load. Peak load testing is needed to avoid the duress scenario.
Our Login VSI product is used by 95% of Desktop Virtualization hardware and software vendors because it allows for setting up a peak load easily through virtual users and synthetic workloads, rapid test cycles and delivery of results, and provides a view into performance from the end user perspective. It’s used in the largest enterprise VDI implementations in the world to measure performance against peak loads and to continuously test the behavior and impact of software and infrastructure updates.
Finally, production code and infrastructure is instrumented as a best practice to understand behavior so that proactive management of system components can occur, avoiding outages. I believe that Login PI provides the same function for production VDI. It provides an early warning system from the end user perspective and will alert you to looming performance issues so that you can proactively address these issues before the help desk lights up.
I believe we can help any VDI implementation be more successful by playing a role in thorough and regular load testing, and also in providing production performance insights from the end user perspective (this blog about the VDI lifecycle explains how to guarantee the best performance in every stage of a VDI environment). Let us help you avoid an outage to your user community due to peak load.