In my last post, I suggested that the Dell PowerEdge R805 server was my current choice to use as our new virtual server. That was correct. However, I happened to be browsing the Dell Outlet on Wednesday and mentioned to a few friends on the #citrt IRC channel that I was doing so. A couple of them started looking along with me and pointed out two particularly nice servers at particularly nicer prices! Both were Dell PowerEdge 1950 III servers, which are rack-mount 1U servers (makes them small and easy to rack). Both have two Intel Core 2 Quad 2.33GHz processors and 16GB RAM in 4 DIMMs, both include 3 year warranties, and one of the two came with a DRAC (Dell’s remote access card), four 73 GB 15,000 RPM SAS hard drives and a RAID controller, and dual power supplies. The other came with an 80GB SATA drive, no RAID controller, and one power supply. The former was of course a bit more expensive than the latter, but together they added only about $300 to the brand-new price I was quoted on the R805 server! This doesn’t include licensing which I’ll cover in a minute. Needless to say, I have purchased these servers and they should arrive soon!
One of my goals for the new system is redundancy. If something goes down, I’d really like a second system around to stay up and running, in particular when using virtualization it does create bigger single points of failure (one phyical server going down takes down multiple virtual servers), which is one of its biggest weaknesses. This can be mitigated by using shared storage (hence the SAN) and multiple servers that can take on the virtual machines the “down” physical machine can’t run temporarily, even if it runs a bit slow from the additional load (even better if your secondary server is not heavily loaded!).
Having the R805 would be great, but my next-best server is a Dell PowerEdge 1800 that’s three years old, with a single Xeon 3.0GHz processor and 6GB of RAM. It’s a very nice server, but it wouldn’t be able to shoulder a load the R805 could handle easily so it would only be able to run absolutely critical machines. Additionally, the Xeon processor is too old to support Intel’s VT (Virtualization Technology) extensions that make running a virtual server hypervisor easier, and allows 64-bit guest operating machines to run. If we make the move to Exchange 2007, there would be no backup server for it to move to, and it would be one of our most critical servers!
VMware’s ESXi hypervisor runs just fine on the PowerEdge 1800 (as long as you don’t need 64-bit guests); I have it running now with a couple of VMs and it’s barely breaking a sweat. It will work even better on the nice “new” PowerEdge 1950’s, but the other thing that having VT-capable processors will help me with hypervisor selection options. If I don’t end up running the now-free VMware ESXi, which I’ve tried and like but keeps you from using some of the really cool features unless you buy their Virtual Infrastructure/Virtual Center packages (these easily get into the same price as the hardware for both servers I bought!). Microsoft’s new Hyper-V virtualization platform (separate entirely from the old Microsoft Virtual Server product) is capable of running only on processors supporting VT, and now I’ll have two of them, the magic redundant number. It does preclude using the PowerEdge 1800 as an third backup, but down the road it will offer some of the similar moving of “live, running” virtual machines from host to host, and Microsoft is releasing their Virtual Machine Manager (VMM) soon as well for management, which will be siginificantly more cost-effective for us given charity pricing than VMware’s Virtual Center. But some of this stuff isn’t going to be here right away, and VMware’s solutions aren’t necessarily lacking anything we desperately need. What’s the answer? For now, it’s keep researching, maybe even try both when the servers arrive, and see which is the best fit for us. Microsoft certainly wants to take over the market and they’ve been successful in other areas in doing so after entering late, but it’s way too early to tell in this case, in my uneducated opinion! I do think VMware will be around for a while, and is not a poor choice from a longevity perspective yet.
Licensing is the only “kicker” with my new servers. Microsoft licenses (that link has a cool calculator, but it gives retail and not charity prices!) their Microsoft Server Datacenter software per socket (physical processor) in each server. So if you have a dual-processor system, you buy two licenses. The R805 is a dual-processor system, but the 1950s are as well, and I just doubled my processor count and therefore licensing cost! That’s OK, I’m likely going to, for now, use the Datacenter license from the PE 1800 for one processor in one of the new servers and use a Server Standard license on the PE 1800 for now that I’ll be freeing up by virtualizing to pre-licensed VMs. This will just add around $500 to the cost of getting a second server, making the hardware and licensing increases in going from one to two servers come in right at $800. That’s not bad for a second server, doubling your quad-core processing power and RAM as a consequence!
Because these servers were in the Dell Outlet, where items in your cart last only 15 minutes unless you modify them, the server have now been purchased and are estimated to ship on October 1st. I’m still waiting to pull the trigger on the MD3000i SAN and I’m still researching backups (which just had its budget cut a bit with this server swap!). I did hear from my Zones rep that October 1st begins Microsoft’s new fiscal year, and they will be announcing any pricing changes at that time. Pricing on Datacenter could go up, or stay the same (well, it could go down but how likely is that? Exactly!). Since I have the servers for sure now, I may go ahead and grab the Datacenter licensing before the end of the month for that reason if I can (my boss is out of town at the moment; he approved the server purchase earlier waiting to change airplanes at an airport in the Bahamas on a business trip. No, he called me first for some technical assistance, I didn’t bug him until I had him on the phone already. Yes, I should have asked if he needed an assistant for his business trip :-)
So, one more step completed in the process, many remaining. Also, I’m possibly going to need (or want very much to have) a cheap or free server rack that I can pick up locally. Just saying, if you happen to be throwing one away and are nearby :-)
David, we run Microsoft Hyper-V in a Windows Server Core Failover Cluster. I’d be glad to talk about my experiences using the product with you.
Hi Scott. I’ve never done any clustering before but I’m familiar with some of the concepts. I hadn’t actually thought about doing that with my system but it does sound like, with the duplicate hardware we’ll have, that it might be a perfect solution to give us higher redundancy! I’d love to talk to you about it sometime, and I may mention it to a consultant friend of mine I’m supposed to have lunch with today; he could probably set up clustering in his sleep (and may have done so in the past), for multinational corporations :-) Thanks for the tip.
This sounds very cool. I am hoping to move in the direction you described eventually and will have to keep the Dell Outlet in mind when I do.
I just got our MD3000i racked up. Dell support is calling me tomorrow to assist with the setup. I’ll let you know my thoughts. Feel free to email me any questions.
Just wondering about how running VMware’s ESXi hypervisor worked for you.
We have windows server 2003 and I would like to upgrade to 2008. I was going to move it to a box with a xeon 5130 chip.
I am just a little worried about using VMware’s ESXi
What are you worried about specifically? I’m running three servers with ESXi (using the VMware Essentials package that retails for $999 and sells on the street for about $750, it adds a vCenter Server license so you can manage up to three servers from a single interface) and have been for over a year. It’s excellent, but keep in mind with only one server and no SAN you can potentially end up with “too many eggs in one basket” if your ESXi server fails without a second ready to take its place that has access to the virtual machines. If that one server dies, it takes all the servers on it down until it’s back up. That would be the only “gotcha” I’d be aware of. It’s one reason why I have a SAN and at least two servers–should one server die, the other can run all of the VMs they were previously sharing, achieving better uptime (and cost savings since I’m running 23 VMs on primarily two servers) than spreading things out on physical hardware for every server. But you need to get to shared storage plus two servers before you get some of those benefits.
However, even one ESXi box can work really well as long as you keep the limitations in mind and have backup plans, and enough tolerance to survive for a while if the box is down.