This is part 4 of the mini Tech Tip series for vSphere Capacity Planning. In part 1 of this series I covered the importance of using the right metrics and tuning the knobs to assess your capacity risk. In part 2 of this series, I covered which visuals you can use in vCenter Operations to answer key questions around datacenter or cluster wide capacity usage and risk monitoring. In part 3 of this series, I covered insights on right sizing your VMs.
In this 4th Tech Tip, I will cover a simple phased methodology that enterprises can use to optimize your infrastructure utilization and realize savings.
The first step is to decide on your optimization phases and how aggressive you want to be across your environment. E.g.You may choose to optimize heavily in staging/UAT or test-dev but not touch production environment.
Phase 1: Reclaim Waste
Often I see huge sometimes even 20-50% savings that customers can get from simply reclaiming waste from unused ( idle/powered off VMs). Figure below shows widgets from the vC Ops’ dashboard that give you a sense of how efficiently you are using your infrastructure. The top left widget shows a population distribution over time of how efficiently is your infrastructure utilized. Green is good. The widget below that shows distribution of VMs across say a selected cluster. And the view at the bottom right gives you a list of the idle and powered off VMs that you could reclaim unused capacity from. You could also evaluate the Capacity Optimization report and start from there. A great starting point to analyze opportunities to reclaim in your environment.
Phase 2: Increase Utilization
The next phase is around increasing the utilization of your clusters. You could evaluate your under-utilized clusters based on vC Ops stress based analytics. There is a view you could use under planning called Under-used Hosts and Clusters.
Next you could evaluate using vC Ops’ what-if analysis which clusters can you consolidate? You can model -if all the VMs of a source cluster would fit into an existing target cluster. You could also use this almost hidden but extremely useful feature of vC Ops’ that tells you ahead of time how many Small, Medium and Large VMs can fit into your target cluster. The Small, Medium and Large VMs are based on the distribution of your existing VMs on the selected inventory object on the left.
Phase 3: Over-Commit
The next phase is around evaluating how many VMs you can squeeze into each of your clusters WITHOUT impacting performance. This is a difficult but important question and vC Ops shines here. We have the ability to tell you the demand across all the VMs on your cluster. This is the actual resources that the VMs are asking for. We use this to figure out how many VMs we can fit in. This guarantees performance of the VMs since we are accounting for their actual demand. The density widget on the dashboard gives you the actual and potential consolidation and CPU / Memory overcommit ratios you could play around with. Even if you take 50% of what is recommended it could give you a head start and achieve those extra 5% of savings for the year while making sure that the performance of your VMs is not getting impacted.