The DRS Dilemma.
vSphere’s Distributed resource scheduler has long been a part of our data centers, keeping an ever vigilant watch over our hosts and their workloads. Distributed resource scheduler, or DRS for short, is responsible for choosing the best host for each of our virtual machines to run on. When you deploy a new VM, DRS is there for the initial placement. When a cluster is out of balance or a host is running low on resources, DRS is there to move VM’s to other hosts in the cluster. Moreover, when it is time to perform maintenance on a host, DRS is there to evacuate the VM’s and place them on other hosts in the cluster. DRS is one of those things that once you enable it, you just can’t seem to live without.
As awesome as DRS is, however, it is constrained to intra-cluster virtual machine moves. That means when all of the hosts within a cluster have run out of available resources there’s little to nothing that DRS can do. Imagine a Tetris screen that’s completely full. No amount of tinkering with the placement of your pieces can free up space to add more pieces. The only options are to remove or modify pieces to make more room.
DRS and vROps, better together!
Thankfully however, vRealize Operations Manager has the answer. When DRS and vROps work together some interesting possibilities open up. vROps can expand DRS’ capabilities by coordinating virtual machine moves between clusters. This means that workloads can be balanced across an entire data center, further eliminating hot spots and maximizing your hardware investment.
And if you have concerns about mixing prod and dev environments, or if you have special licensing constraints, vROps has you covered there as well. With just a few clicks, you can create your own custom data centers to ensure your VM’s don’t run where they aren’t supposed to. vROps automated workload balancing works by looking at each of your clusters and their virtual machines and calculates the most efficient way to distribute the VM’s across your data center (or custom data center).
Before committing to any moves, vROps shows a before and after comparison of what your environment will look like and what VM’s will be moved to achieve virtual nirvana. If the user accepts the moves, vROps initiates the vMotion(s) by handing the VMs off to their new cluster and letting that cluster’s DRS and storage DRS handle the individual host and datastore placement. Of course, DRS must be set to fully automated for each of your compute clusters in order to take advantage of this great new feature.
Up until this point, DRS has been a reactive tool that moves workloads after contention has occurred. How great would it be if we could leverage vROps to help predict when contention is likely to occur? After all, many production VM’s see workload spikes at regular intervals. For example, when users log on to their applications in the morning, or when backups run at midnight, or when the regularly scheduled database cleanup script kicks off.
Once again, vROps gives DRS super powers by feeding workload statistics to DRS in a feature called Predictive DRS. Predictive DRS, or pDRS identifies workload patterns for each virtual machine in order to predict future resource needs. If a host doesn’t have enough resources to keep up with future demands, pDRS will move workloads to other hosts in the cluster before contention can even occur. vMotions can be taxing on a host’s resources, so why put the additional burden on your hardware when it’s already stressed out? This is a vast improvement compared to reactive approaches to contention.
Enabling pDRS is easy. Log in to your vCenter server and edit your cluster’s DRS settings. Under the DRS automation settings enable predictive DRS and repeat this for any other clusters which you’d like to enable predictive DRS. Next, log in to vROps and go to the administration tab and edit your vSphere solution. Under the advanced settings set the “provide data to vSphere predictive DRS” setting to true.
By combining the automation of DRS with the intelligence of vROps, you’ll be able to break free from firefighter mode and focus on more important things. For example, finding the perfect cup of coffee, or I suppose that project your boss keeps hounding you about. Either way, be sure to check out vROps today by visiting https://www.vmware.com/products/vrealize-operations.html.