[ED. NOTE: This post is written by Sanjay Castelino, SolarWinds' VP of Product Marketing. He has this to say about himself... "I'm a long time technologist and a geek at heart. My path to SolarWinds has taken me through hardware design, software engineering, and finally product marketing."]
By now, you’ve all heard about vSphere5, and if the anecdotal survey during one of the massive sessions at VMworld Las Vegas is any indicator, many of you are thinking about upgrading soon. So I thought it would be interesting to put some of my thoughts out there on the opportunities and challenges that you may encounter during and after your upgrade (or new purchase if you’re just getting started).
First, kudos to VMware for diving into the world of storage. Any of us who’ve looked at or tried to virtualize know that storage has been a major headache in the process. As I listened to the new features in v5, the biggest ones were clearly Storage DRS (Distributed Resource Scheduler), Storage Profiles, and the virtual storage appliance. While I could go on about each of these for some time, let me stick with the first one.
DRS is a great feature. It is a linchpin in fulfilling the promise of virtualization because it helps to ensure that you don’t starve your workloads in the effort to squeeze every ounce of efficiency out of your hardware. With v5, DRS now accounts for storage capacity and latency at the datastore level – those last two words are important – let me explain.
The hypervisor still can’t see past the datastore, but datastores are made up of LUNs, and LUNs are going to belong to specific RAID groups, and ultimately disks that are connected to your network via controllers. You can still set up your storage such that multiple datastores are connected to the same RAID group, or so that all your datastores in a datastore cluster use disks that flow through a single controller. However, if you don’t manage this setup well, you will end up with DRS recommending vMotions that won’t improve performance because two datastores on a cluster may still be going through the same physical bottleneck. One tool that VMware is providing to ensure that your vMotion workloads to the right type of storage is profiles. Profiles allow you to categorize your storage with what amounts to metadata, and once you have your storage in a profile you simply make sure that your workloads stay in the right profiles. But you still have to do the work to figure out what storage gets mapped to what profile.
So what’s a VM admin to do? Well, two things.
1. Make sure you know how your storage is set up today. If you’re already using vSphere,how do you have datastores set up and how are they mapped to your workloads?
2. Once you think you’ve got things set up right and DRS is making recommendations on workload movement, you can use a product like SolarWinds Storage Manager, Powered by Profiler, to verify that the workload vMotions would in fact be beneficial.
There’s a lot more to explore on vSphere5 storage features, so expect to see more written on this topic in the near future.