Q&A

SAN, Saving: Dell and Virtualization Come to Scott County

Dell Compellent SAN delivers agility, ease of use to Scott County, Minn. and it could save them $60K/hour in the event of a disaster.

• By Bruce Hoard
• 11/01/2011

Virtualization Review Editor in Chief Bruce Hoard recently interviewed Todd Croy, IT manager for Scott County, Minn., about his selection and deployment of a Compellent SAN from Dell Inc.

VR: What were the primary pain points that led to your purchase of a Dell Compellent SAN?
Croy: When I started in 2005, the county had just begun virtualization with VMware in a limited capacity. They were running VMware on a local server and those virtual machines [VMs] were on the local server. The problem was [that] at that time there was no way to back up the virtual instances easily. They did look at another product, a software product, to replicate that data somewhere else. So when I started here, one of my first tasks was to look for a NAS solution. We thought that was what we needed to store these VMs on, or at least to replicate the data to a NAS solution. As we started looking at what our needs were and how we were going to use the solution, we discovered that we needed a SAN. One of the cities in our county that had looked at Compellent prior to us gave us the idea, so we started looking at the Compellent and what it could do as far as replicating those virtual instances to another disaster recovery site, and that gave us that backup we were looking for. That was the real big pain point.

VR: What was it about the Dell Compellent SAN that attracted you? What features sold you on it?
Croy: We looked at a solution from IBM, and we also looked at a NetApp solution. Those were the two that were within our budget. What actually sold us on the Compellent was its ability to handle the kind of data we were going to be throwing at it, and its ease of use. We wanted to make sure we had an easy solution for going in there and making changes as needed. Those are the two big selling factors—along with the replication, of course—and the costs of the features that the Compellent provided versus those from NetApp, where you have to purchase additional modules.

VR: Let's take a look at some of those features and functions, starting out with the Storage Center.
Croy: There are separate agents within the NetApp that you have to purchase for replication and thin provisioning. I believe those were separate modules after you actually purchased the appliance. That cost was prohibitive for us.

VR: Tell me about the array-based snapshots.
Croy: The solution we purchased with Compellent included two systems. We purchased a main system for our main datacenter, and then we have a second system at our disaster recovery [DR] site connected by private fiber that we own. We replicate all the data from our primary site to our DR and vice versa, and then part of our backup strategy is to take nightly snapshots of those LUNs on the primary system or wherever that data is live and replicate it to the other system. That gives us a couple of things. Obviously, it gives us a point-in-time copy of those VMs to go back to for restore or disaster purposes, and it also gives us the ability to do some testing. I can take a server—take one of those snapshots—restore it in its own environment and do updates or patching, or whatever I need to do with that machine without impacting production. Those are the main things that we use the SAN for, other than just holding our data. We just went through a refresh of our

Compellent systems, probably six to eight months ago, and in that refresh, we purchased a feature called Live Volume. Live Volume allows us to move our storage in the same way vMotion works in VMware. We can move that storage between SANs, and that allows us to run production servers in either of our locations seamlessly. It also allows us to move them between locations without any downtime.

VR: Where are your datacenter and the other remote location?
Croy: The remote location is 17 miles from our primary location.

VR: Talk to me about storage virtualization.
Croy: Virtualization has the basic ability to provide thin provisioning live with vMotion. Moving that storage back and forth between the two sites is a huge win for us. I'll give you an example. We have a SQL cluster, and one node of it is in our primary datacenter, while a secondary node is in our DR. They both connect to the primary SAN in our government center, but we have the ability to move that cluster to the DR node within 30 minutes. Once the Compellent sees that it's actually using the node at the DR site, it figures out that it would make more sense to run locally at the DR site, so it switches the data over to the DR SAN.

VR: As far as implementing the SAN went, how did that work between them coming in and doing the work versus you doing it internally? Did they work with you? How did that go?
Croy: We worked with the reseller to purchase the SAN, and once it was purchased, the reseller came in and set it up for us. There's a lot to setting it up—not that it's hard, necessarily, you just have to make sure that everything goes where it needs to go. We had full support from the reseller and I honestly wish we had more vendors that paid as much attention as [Dell] and the reseller to make sure things are working right.

VR: Can you talk about how much money you're saving?
Croy: Since we didn't have a SAN previously, it's hard to say what the true savings are. I know we have savings in the amount of time our staff has to spend re-provisioning servers or doing backups and restores from tape. Obviously there's the benefit of having that kind of insurance, of having that disaster site up and available. It's a cost to implement, but once you have it, it's a necessity that we really needed to have.

One way to evaluate the savings is by determining what the cost would be in the event of a disaster. You look at some of these figures from disasters these days where it took 10 to 12 days for an organization to get back up and running. For data that's running on the SAN and in our environment where 95 to 98 percent of our environment is virtualized on the Compellent, we're looking at—knock on wood—I'm saying it would take us less than a day to get back up and running again.

I think we figured that if we have a complete outage with staff time in the building, we are out somewhere in the neighborhood of $60,000 an hour.

VR: Have you had any outages yet?
Croy: None to speak of. No big disasters, just a couple of power outages here and there, but the datacenter has always been protected, so we haven't really had to roll over there.

VR: Are there any particular
lessons that you've learned that you could pass along to someone else who might be in the situation you were in?

Croy: You have several technologies that you're working with that you really need to understand in terms of how they work and play together. Everything that we've done for the most part has been in-house and kind of learning on the job, and in some areas we just took the plunge and saw how things were going to go. It's making sure you're doing best practices, and setting up things for the best performance and the worst-case scenario where you have to restore data.

A lot of people today like to put several VMs into one LUN. We like to split them up. We like for each VM to have its own LUN. It's on its own, it has its own restore point. It gives us a lot more flexibility to do what we do without affecting anything else.