Moudry: When Hurricane Ike came, we were using another DR support company, and they gave us facilities to recover our data. They were also doing our backups.
We went to that site to recover systems, and we had a hard time recovering anything. We were testing it, and it was really cumbersome. We tried to get servers up and running. We stayed there to recover one whole day and never got even a data center recovered.
So William and I were chatting and thinking that there's got to be a better way. That’s when we started testing a lot of the other virtualization software. We came to VMware, and it was just so easy to deploy.
We made a proposal to our executive committee, and it was an easy sell. We did the whole project for the price of one year of our old DR system.
Gardner: William, what were your top concerns about change?
Chambers: Our top concerns were just avoiding what happened during Ike. In the building we're in in Houston, we were without power for about a week. So that was the number one cause for virtualization.
Number two was just the amount of hardware. Somebody actually called us and said, "Can you take these servers somewhere else and plug them in and make them run?" Our response was no.
That was the lead into virtualization. If we wanted everything to be mobile like that, we had to go with a different route.
Then, once you get into virtualization, you think, "Well, okay, this is going to make us mobile, and we'll be able to recover somewhere else quicker," but then you start seeing other features that you can use that would benefit what you are doing at smaller physical size. It's just the mobility of the data itself, if you’ve got storage in place that will do it for you. Recovery times were cut down to nothing.
Simpler to manage
There was ease of backups, everything that you have to do on a daily maintenance schedule. It just made everything simpler to manage, faster to manage, and so on.
Gardner: And so for you as an SMB with 200 employees, what requirements were involved? You obviously don't have unlimited resources and you don't have a huge IT staff.
Chambers: It’s probably what any other IT shop wants. They want stability, up-time, manageability, and flexibility. That’s what any IT shop would want, but we're a small shop. So we had to do that with fewer resources than some of the bigger Exxons and stuff like that.
Moudry: And it can't cost an arm and leg either. We're an insurance broker. We're not a carrier. We are between the carriers and agents. With our people being on the phone, up-time is essential, because they're on the phone quoting all the time. That means if we can’t answer our phones, the insurance agent down the street is going to go pick up the phone, and they're going to get the business somewhere else.
Also, we do have claims. We don't process all claims, but we do some claims, mainly for our stuff that's on the coast. After a hurricane, that’s when people are going to want that.
We have to be up all the time. When a disaster strikes, they are going to say, "I need to get my policy," and then they are going to want to go to our website to download that policy, and we have to be up.
Gardner: Why did you go 100 percent virtualized in such a short time?
|Cambers and Moudry|
Chambers: We did that because we’ve got applications running on our servers, things like rating applications, emails, our core applications. A while back, we separated the data volumes from the physical server itself. So the data volume is stored on a storage area network (SAN) that we get through an iSCSI.
That made it so easy for us to do a physical-to-virtual (P2V) conversion on the physical server. Then in the evenings, during our maintenance period, we shut that physical server down and brought up the virtual connected to the SAN one, and we were good. That’s how we got through it so quickly.
Moudry: William moved us to VMware first, and then after we saw how VMware worked so well, we tried out VMware View and it was just a no-brainer, because of the issues that we had before with Citrix and because of the way Citrix works. One session affects all the others. That’s where VMware shines, because everybody is on their independent session.
Gardner: Where are your data centers?
Moving to colos
Moudry: Right now it’s Houston and San Antonio, but we are moving all of our equipment to colos, and we are going to be in Phoenix and Houston.
Gardner: So that’s even another layer of protection, wider geographic spread, and just reducing your risk in general. Let’s take a moment and look at what you’ve done and see in a bit more detail what it’s gotten for you. Return on investment (ROI), do you have any sense, having gone through this, what you are doing now that perhaps covered the cost of doing it in the first place?
Moudry: We spent about $350,000 a year in our past DR solution. We didn’t renew that, and the VMware DR paid for itself in the year.
We're working with automation. We're getting less of a footprint for our employees. You just don’t hire as many.
And we are not buying equipment like we used to. We had 70 servers and four racks. It compressed down to one rack. How many blades are we running, William?
Chambers: We're running 12 blades, and the per year maintenance cost on every server that we had compared to what we have now is 10 percent now of what it was.
Gardner: I notice that you're also a Microsoft shop. Did you look at their virtualization or DR? How come you didn’t go with Microsoft?
Chambers: We looked at one of their products first. We've used the Virtual PC and Virtual Server products. Once you start looking at and evaluating theirs, it’s a little more difficult setup. It runs well, but at that time, I believe it was 2008, they didn’t have anything like the vCenter Site Recovery Manager (SRM) that I could find. It was a bit slower. All around, the product just wasn’t as good as the VMware product was.
Moudry: I remember when William was loading it. I think he spent probably about 30 days loading Microsoft and he got a couple of machines running on it. It was probably about two or three machines on each host. I thought, "Man, this is pretty cool." But then he downloaded the free version of VMware and tried the same thing on that. We got it up in two or three days?
Chambers: I think it was three days to get the host loaded and then re-center all the products, and then it was great.
Moudry: Then he said that it was a little bit more expensive, but then we weighed out all the cost of all the hardware that we were going to have to spend with Microsoft. He loaded the VMware and he put about 10 VMs on one host.
It was running great. It was awesome. I couldn’t believe that that we could get that much performance from one machine. You'd think that running 10 servers, you would get the most performance. I couldn’t believe that those 10 servers were running just as fast on one server that they did on 10.
Chambers: That was another key benefit. The footprint of ESXi was somewhat smaller than a Microsoft.
Moudry: It used the memory so much more efficiently.
Gardner: You mentioned vSphere, vCenter Site Recovery Manager, and View. Is that it? Are you up to the latest versions of those? What do you actually have in place and running?
Chambers: We have both in production right now, vCenter 4.1, and vCenter 5.0. We’re migrating from 4.1 to 5.0. Instead of doing the traditional in-place upgrade, we’ve got it set up to take a couple of hosts out of the production environment, build them new from scratch, and then just migrate VMs to it in the server environment.
It's the same thing with the View environment. We’ve got enough hosts so we can take a couple out, build the new environment, and then just start migrating users to it.
It all happened much quicker than we thought. Once we did a few of the conversions, of the physical servers that we had, and it went by so fast that it just happened that way. We were ahead of schedule on our time-frames and ahead on all of our budget numbers. Once we got everything in our physical production environment virtualized, then we could start building new virtual servers to replace the ones that we had converted, just for better performance.
We were able to do it without disruption, and that was one of the better things that happened. We could convert a physical server during the day, while people were still using it, or create that VM for it. Then, at night, we took the physical down and brought the virtual up, and they never knew it.
Gardner: How about some other metrics of success?
Copying the template
Moudry: Making new servers is nothing. William has a template. He just copies it and renames it.
Chambers: The deployment of new ones is 20 minutes. Then, we’ve got our development people who come down and say, "I need a server just like the production server to do some testing on before we move that into production." That takes 10 minutes. All I have to do is clone that production server and set it up for them to use for development. It’s so fast and easy that they can get their work done much quicker.
Moudry: Rather than loading the Windows disk and having to load a server and get it all patched up.
Chambers: It gives you a like environment. In the past, where they tested on a test server you built, that’s not exactly the same as the production server. They could have bugs that they didn’t even know about yet, and that just cuts down on the development time just a lot.
Gardner: Any advice for folks who are looking at the same type of direction, higher virtualization, gaining the benefits of DR’s result and then perhaps having more of that agility and flexibility? What might you have learned in hindsight that you could share with some other folks?
Chambers: If you are going to use virtualization, then get in and start using it on a small basis. Just to do a proof of concept, check performance, do all the due diligence that you need, and get into it. It will really pay off in the end.
Moudry: Have a change control system that monitors what you change. When we first went over there, William was testing out the VMs, and I couldn’t believe, as I was saying earlier, how fast it is. We have people who are on the phones. They're quoting insurance. They have to have the speed. If it hesitates, and that customer on the phone takes longer to give our people the information and our people has hard time quoting it, we’re going to lose the business.
When William put some of these packages over to the VM software, and it was not only running as fast, but it was running faster on the VM than it was on a hard box. I couldn’t believe it. I couldn’t believe how fast it was.
Chambers: And there was another thing that we saw. We’ve got a lot of people working at home now, just because of the View environment and things like that. I think we’ve kind of neglected our inside people, because they'd rather work in a View environment, because it's so much faster than sitting on a local desktop.
Moudry: When somebody works at home, they're at lightning speeds. Upstairs is a ghost town now, because everybody wants to work from home. That’s part of our DR also. The model is, "We have a disaster here. You go work from home." That means we don’t have to put people into offices anywhere, and with the Voice over IP, it's like their call-center. They just call from home.
Chambers: They can work from different devices now, too. I know we’ve got laptops out there, iPads, different type of mobile devices, and it's all secure.