Help choosing a virtualization solution

September 21, 2012 8:54 AM

Choosing between virtualization solutions for server migration (VMware & Microsoft Hyper-V) when we know nothing about either product?

The trick is nobody on staff knows squat about production virtualization products. Many of us have worked with Remote PC, VMware on an individual basis, Parallels, but nothing on this scale. So what we're all wondering is: How the heck do we compare two different products, most likely with a fair difference in price, that we have little true knowledge of? We don't even know what to look for or what questions to ask. Are there specific Virtualization questions we need to ask to make sure we're not going to look back 2 years from now and realize "Crap, we went with the wrong solution for our needs!!!" Anyone worked with both and know first-hand what to be looking for?

A little background:

We're in the process of working with 2 vendors on upgrading our servers- about 10 spanning across 2 physical locations (DNS, DHCP, AV, Constituent database, file share, Financial DB, etc). They're all 10+ years old and run server 2k3 or 2000. To say they're on their last leg is kind. We also have about 10 remote PCs set up because the database software doesn't run on their OS's, so they remote in to run the database (maybe not the best solution logistically, but works well enough).

We're finally getting funding to upgrade our aging infrastructure, and the 2 vendors are all about virtualizing our servers and remote stations onto a virtualized solution with one of two products - 1 is all about VMWare and 2 is Microsoft Hyper-V with the promise that Microsoft will offer substantial discounts to our type of nonprofit (cost is a factor!), all connected to a SAN.

Things I'm primarily concerned about:
1) Ease of backup & redundancy (I realize the 2, while they can work in conjunction with each other, are not the same). I'm concerned with "putting all our eggs in one basket." Both companies have said there's be 2 servers that replicate to each-other, basically as a fail-over. Almost like a raid-1. Is this really effective? Any pitfalls with either company?

2) Ease to deploy new virtualized servers or workstations when needed. I'm very much a jack of all trades and don't want to spend 2 hours reminding myself how to do ir correctly every time.

3) It. Just. Works. Everyone here has a lot of other duties, and none of us plan on (or have the time to) become Virtualized Experts. We don't want a product that has weird quirks and maybe kinda works half the time.

Things the higher-ups are concerned about? Cost.

I realize a lot of these questions are ones I should pose to the vendors- however, we've dealt with both in the past and no matter the solution in prior cases, their answer is "OF COURSE IT WORKS!!!" even if it turns out it always doesn't.

posted by MeProxy to computers & internet (8 answers total)

I'm a bit biased having worked with two very large VMWare deployments in different companies. Rather than try to sell you on VMWare specifically, I'd advise you to define your project budget and contact some vendors who are willing to show you both products (probably different vendors, necessarily) in depth, and talk about your specific concerns. Some really big mis-steps can be made with either option, like not properly planning for hardware failover, which they can review with you.

Your specific need to deploy not only servers but workstations is cause for some very in-depth conversations about what that looks like.
posted by odinsdream at 8:57 AM on September 21, 2012

By the way, if you want any nitty-gritty details about VMWare from a real user just memail me when you get to that stage.
posted by odinsdream at 9:00 AM on September 21, 2012

Regardless of the product that you choose, you will need to learn about virtualization. It is up to you to build images for deployment, and for Hyper-V (the one I know) this involves some ramp-up in terms of learning how to make a generic base image. Probably VMWare has a similar curve.

Have you considered cloud solutions, such as Amazon EC2? At least that would save you the trouble of failover and redundancy.
posted by crazycanuck at 9:18 AM on September 21, 2012 [1 favorite]

I am currently overseeing our companies server consolidation from about 15 servers into a Virtual Machine cluster. I think we had a very similar circumstances, but we're not a non-profit.

Initially we reviewed the solutions by Microsoft, VMware, and Xen (opensource). We eventually decided on VMware as we want something with commercial support and wanted to avoid Microsoft (for reasons I wont go into).

At that point we contacted and talked to a couple vendors to see what they could do for us. After a few sales meetings, we ended up settling on Dell who sold us the entire package of servers, software licenses (Windows server, VMware, VCenter, and AppAssure), networking and a SAN. I do believe having 1 vendor made this very easy for us.

After all the parts arrived Dell engineers spent many,many hours supporting/educating us on installation, configuration, and best practices. The Dell staff reminded me of the experiences I had with my customers when I was a System Engineer with SGI. They didn't know everything, but it was very much worth while.

We're finishing up now, still need to do some electrical work in our server room and getting AppAssure (backups) running. But we've migrated several hosts to the new cluster and we're very happy with the result.

That was my experience. If you want a recommendation, I'd call Dell and chat with their sales/engineering team and see what they can do for you, its not a hard sell or anything. Also I know they offer Microsoft's Hyper-V if that's appealing to you.

Good Luck!
posted by rickim at 9:53 AM on September 21, 2012

3) It. Just. Works. Everyone here has a lot of other duties, and none of us plan on (or have the time to) become Virtualized Experts. We don't want a product that has weird quirks and maybe kinda works half the time.

I have never heard "It. Just. Works." applied to any enterprise software, anywhere. Even dead-simple CRM. "It. Just. Works." is something you can expect from iOS and pretty much nothing else.

Whatever solution you go with, you need a virtualization expert, and your higher-ups need ot understand this. It's a lesson you can proactively learn now, or can learn in a few months after your fourth major outage in a month.

Talk to your leadership and insist that at least one of you get serious training in the software and get some duties taken off your plate so you can be the go-to guy. Sounds like a good chance to get some career-enhancing training.
posted by Tehhund at 10:11 AM on September 21, 2012

Microsoft definitely does offer some attractive discounts to nonprofits (on pretty much all of their software). It's one thing they've always been really great about.

In a pure "specs" contest, the new Hyper-V that came out with Server 2012 is more powerful than the comparable VMware solution, though I doubt you'll reach the upper limits of either platform. But the new Hyper-V does have one really useful feature that bears mentioning: Replica, which makes VM backup easier and cheaper than it's ever been before.

At the end of the day, both solutions should be able to supply everything you need (and a lot more) if they're engineered properly. So I'd say what you should really be looking at is which solution provider you're most comfortable with and what your final costs will be (including service fees).

Also, I'm sad to say that there's really no such thing as "It just works" in a complex, multi-server environment. You can get to "It just works... almost all of the time," but someone (even if it's a contractor) will have to be available to do occasional maintenance. That's true regardless of whether or not you go virtual, cloud, or whatever else.

The only situation in which you don't have to worry about any server maintenance is if you go with a complete suite of software-as-a-service (SaaS) solutions. If you're not familiar with them, they're products like Salesforce, QuickBooks Online, Gmail, etc. that run entirely "in the cloud" (i.e. on servers in someone else's data center) that you access with a Web browser.
posted by GnomeChompsky at 10:16 AM on September 21, 2012

The trick is nobody on staff knows squat about production virtualization products.

(It seems to me that in the SMB space, virtualization is mainly a sneaky way of replacing equipment and operating costs with labor and licensing costs. Mid-level sysadmins who can run 15 servers are a dime a dozen, but once everything is virtualized you need an expensive senior network admin, an expensive virtualization admin, and a fantastically expensive storage infrastructure admin....not to mention a good friend in Finance to manage your astronomical vendor support contracts.

Note: this is great for you if you plan to become one of those highly-paid virtualization experts!)

That said - at the scale you're operating at, in my experience, platform and vendor don't matter nearly as much as a good understanding of network and storage infrastructure. Most sysadmins who manage small networks have never really needed (or been able!) to probe the performance limits of their environment. Those performance limits sneak up real fast, though, when you're running 12 systems off the same disk array, or you're doing all of your storage access over the same switches as the user LAN. Just because you dropped a quarter million dollars on your SAN doesn't mean you'll automatically get better performance out of it than your 6 year old standalone fileserver...

Not to mention how gruesome failures can get without a -lot- of redundancy in place. In a basic SMB environment, if a core switch fails, assuming you have a comparable spare handy, you do a quick repatch, dump on a config that you hopefully have handy or rebuild it from scratch, and then everything springs back to life. In a thoughtlessly virtualized environment, you would do the same, and then have to start manually reviving a bunch of virtual servers that all just died horribly when their storage disappeared.

So: unless you're planning to spend a lot of time getting up to speed, you need to work with a vendor that you trust, prepare to pay them vast sums for support, and ask a lot of hard questions about performance bottlenecks and disaster recovery scenarios.

And now, back to writing that virtualization budget proposal for 2013...
posted by The Prawn Reproach at 11:14 AM on September 21, 2012

Thanks for the suggestions. Strong points taken about the potential staffing issue. Definitely something to bring up, both with vendors and the higher ups. On the bright side, it could give me a distinct career path if I go with it strong!

I also hear the points about "it just works" doesn't really apply to my situation, and I understand that better now. I guess here's how I look at it: years ago, we had an in-house mail server. It was a daily struggle of managing blacklists, why are we on a spam list today, what happened to a user's e-mail, etc? Now we're all through Google Apps for Business and e-mail life couldn't be better. It still requires some maintenance on my end with user accounts, the occasional WTF e-mail, etc- but our lives are much easier for it. I guess that's the angle I was going at with a potential Hyper V vs. VMware solution.
Either way, it's obvious that going this route will require a lot of work from somebody. Also evident we really need to work with the vendors to figure out exactly what extra costs could come into play and how much extra man hours it could cost during and post implementation. Going in, I foresaw virtualizing our servers onto one big platform as a time saver for IT. Less machines == less work, right? Sounds like I'm wrong and need to figure this point out bigtime.

Rambling on, here's a major issue that I don't know how to combat: In dealing with all sorts of vendors in the past, they always make it sounds so easy. Evidence is showing to the contrary. Any tricks to getting honest answers or seeing thru the BS?
posted by MeProxy at 8:11 PM on September 21, 2012