Do you remember Aesop’s fable of the wolf in sheep’s clothing? One day a wolf found the skin of a sheep that had been thrown aside and put it over himself to blend in with the sheep. By doing so, he was able to fool the sheep into believing he was one of them – and enjoyed hearty meals because of it.
When it comes to workload placement, the wolf in our scenario is VMware DRS (Distributed Resource Scheduler). VMware positions this load balancing tool as the only solution organizations need to determine and manage workload placements. This does their customers a disservice. Why? One of the fundamental elements of a well managed virtual environment is having workloads in the right place. Low density is often the result of poorly planned Virtual Machine (VM) placements. The right VM placements, on the other hand, ensure that VMs fit together on physical hardware in such a way that makes the best possible use of capacity resources. Well placed workloads also ensure that enough resource is available for the projected workload requirements thereby significantly reducing risk, the need for motioning and volatility in an environment. When done right, placements are determined by looking at all the relevant constraints such as historical utilization, technical configurations, and policy requirements such as SLA requirements or business policies. The net effect is VM placements that maximize density while giving the workloads the right amount of resource – neither too much nor too little.
VMware DRS, on the other hand, is basically a load balancing tool that looks at short term utilization history to move VMs in response to problems. But just because workloads can be moved around during their busiest times to address shortfalls doesn’t mean this is a good workload management strategy and it doesn’t make DRS a workload placement solution. Reliance on DRS makes organizations reactive rather than proactive. It can increase volatility and risk in an environment because there isn’t a focus on making the right choices in the first place and forecasting requirements. And most importantly, it is not designed to help you figure out how to maximize density.
The moral of the story: VMware may tell you that DRS is a workload placement solution but if you look into the heart of the tool it really is just a load balancing solution. Virtual infrastructure requires real analytics like Cirba’s that consider historical requirements, configurations, policies in order to determine optimal workload placements that both maximize efficiency and minimize risk.
Some organizations that like the idea of a safety net and real-time motioning may choose to run DRS to protect against unforeseen events while using Cirba to plan workload placements and optimize infrastructure. To learn more about workload placement & DRS, watch this short video featuring Andrew Hillier, Cirba’s CTO & co-founder: