In case you missed my last post, I answered the question “What is vRealize Network Insight (vRNI)?” In this post, I want to cover a NEW feature in the NEW release of vRealize Operations – vROps version 6.4. That new feature is predictive distributed resource scheduler. Now it’s likely that you already use VMware’s DRS in your vSphere infrastructure (most vSphere admins do) and it’s likely that you love DRS (again, most vSphere admins do). You may think that DRS is already perfect (and I agree, it’s pretty cool) but what if DRS has the information it needed to actually predict the needs of your applications? With that, your applications can achieve the fastest performance possible and your vSphere infrastructure can have the fastest workload balancing ever. Trust me – it’s really cool! Read on…
What Can DRS and vROps Provide Prior to vROps 6.4?
Some of you reading this may be having some deja vu and thinking “haven’t I heard of vROps and DRS working together before?” The answer is, yes, vSphere and DRS working together isn’t totally new. With vRealize Operations 6.2, Intelligent Workload Placement with DRS was announced. With intelligent workload placement, vRealize Operations helped to rebalance VM workloads across clusters. With intelligent workload placement already in place, you may be thinking that DRS doesn’t need any help balancing workloads within a cluster. After all, DRS already does its job very well but let’s look at the different approaches that DRS has available to it.
Traditionally, DRS has worked by reactively balancing workloads when applications (running on VMs) aren’t getting the resources that they need. DRS analyzes the resources that are allocated (CPU and memory), looks to see if those resources are continuously available, and, if they aren’t moves that application / VM to a host that has those resources available. While this works well in most cases, there is the potential for for contention and, thus, negative application performance impact (albeit very very short). The benefit to DRS is that it’s completely built in with vSphere Enterprise Plus.
The next step up from reactive is balance, which is the intelligent workload placement approach that was discussed above. With the balance method, vSphere DRS and vRealize Operations begin working together. This is an improvement over reactive because vROps is working at a higher level than DRS to analyze performance across multiple clusters. While this is better than DRS, by itself, it also has high overhead and doesn’t do anything to prevent application performance issues and contention before it happens.
What’s New With vRealize Operations 6.4 and pDRS?
Neither reactive or balanced did anything to absolutely prevent application performance issues. While the reactive and balance methods typically do prevent application performance issues, because they tend to work quickly, there’s no guarantee and there’s no prevention of problems, before they happen.
With vSphere 6.5 and vRealize Operations 6.4 (joined in vSphere with Operations Management Enterprise Plus) you can now make the jump to predictive DRS. What pDRS does sounds simple (usually the most elegant solutions are). Predictive DRS predicts application / VM performance problems before they happen and takes action! With minimal overhead, pDRS predicts what resources each VM will need today using the real, past resource consumption of each virtual machine. It combines that prediction with the knowledge of what virtual machines are on each host and the resources that each host is able to contribute. vRealize Operations makes these predictions using its dynamic thresholds. The great thing about Dynamic Thresholds is that they are tailored to each individual VM and application.
Each hour, predictive DRS will take the knowledge that it has of the environment combined with the historical resources utilization of each virtual machine then take action ensure that the virtual machines receive the resources they need BEFORE THEY NEED THEM. The end result is that your critical applications never get the chance to suffer any kind of performance impact.
To be clear, in order to leverage pDRS you need both the new vRealize Operations 6.4 and vSphere 6.5.
You’ll find more information on what’s new in vRealize Operations 6.4 in the slidedeck, below.
I’m starting to upgrade my infrastructure to vSphere 6.5 and vROps 6.4 now, and I recommend that you begin planning your upgrade too!
