To V or not to V(irtualize)…
I know the title is hokey and quirky, but the question itself is something that every organization should be asking itself at some point. “Should we virtualize our servers?”
There are a ton of whitepapers, blog & forum posts, and articles out there, a quick Google search gave me nearly 2.5 million hits (see for yourself Lots of options here!).
I’ll break it down 5 parts… What are the benefits; What are the risks; What flavor should I choose; What should I virtualize; What SHOULDN’T I virtualize.
Part 1: What are the benefits?
This part of the article is where the sales guys go nuts spouting off all kinds of lingo and jargon like “Green Computing”, “ROI”, “Consolidation”, “TCO”, “Capacity Planning”, “Utilization”, “High Availability”, “DR”, and the like. Let me be clear here … there is real validity to a lot of what they saying but like the old saying goes”There are three sides to every story – Yours, Mine, and The Truth”.
The sales guys (Yours) want us all to think that we will suddenly stop using energy (Green Computing), show an immediate and tangible benefit of the dollars we spent on the products (ROI), move from a huge datacenter to a single rack with two servers and a small shared storage(Consolidation), that can be managed by a really smart simian that will require no maintenance (TCO), that will be forever expandable with no more need to spend money (Capacity Planning), that will harness the power of the systems you already have (Utilization), that will be available no matter if interstellar war breaks out and your systems are at ground zero of an alien death ray that melts Podunk Idaho to a sheet of irradiated glass (high availability), and even if the systems are destroyed by said death ray, will be restored and operational in 30 seconds (DR)… WHEW, talk about a run on sentence. This, my friends, is the bill of goods that you can buy (if you do, I have a bridge in Brooklyn that I can sell you REAL CHEAP…I just have to get my Nigerian Prince benefactor to sign the darn thing over).
The IT Guy (Mine) wants to believe the sales guy completely but is smart enough to consult this great new invention … The Internet (thanks for this Mr. Gore). That same Google Search I linked earlier tells a different story than the Sales Guy. The IT Guy, let’s call him Moss (he can be seen here; Moss is AWESOME!!!! And he Likes Milk), may or may not drink the Kool Aid, but he basically thinks the sales guy is full of Vista (Yes I think Vista is the worst thing since… well THIS*).
* InterWeb geeks will get this one
Now The Truth is a little more mundane. Yes virtualizing can save money on power and resource utilization but it depends on how much homework you do and what you are trying to virtualize, what kind of hardware you are running the virtualization on etc. You can realize a Return On your Investment (ROI) but you have to look for it because you have to invest quite a bit of $$$$$ to implement Virtualization correctly (even if you go with free versions of virtualization – we’ll talk about this in a different section). You have to invest $$$$$ in hardware, software, training, etc. to go virtual and see a real tangible results. As part of the last sentence I mentioned training; reality is that it takes a lot more than the smart simian the sales guy talked about to manage your virtualization environment otherwise you have this; Simian IT Guys …
To be clear; there are a lot of real tangible benefits to be had by virtualizing your computing environment. I spoke about ROI earlier and said I would touch on it later … well it’s later. There are tons of ways to virtualize; Microsoft, VMWare, and Oracle all have different ways to do this. Some of them are free (well all of them actually are free in a way – I’ll talk about this in the Flavors installment) and some of them cost a pretty penny. Microsoft has its Hosted Hypervisor called HyperV, VMWare has several flavors but we’ll address only its native hypervisor called ESXi (which is free but limited…but can be VERY feature rich if you pay for it), Oracle has a hosted hypervisor called Virtual Box which is free but only available as an application not a full hypervisor**.
** What is a Hypervisor you ask? Glad you did! Click here for the Lowdown
a. You can run a bunch of virtual servers on a few physical pieces of hardware. For example I have two of these Mamma Jamma Servers(with lots of RAM which is cheap now) and this SAN/NAS solution that are running 20 virtual servers and they have capacity to run 20-30 more. This all takes up 6U’s in a Telecommunications Rack instead of more than 30U’s (Full Rack & consolidated rack) …. What does this mean? Well you use A LOT less power relatively speaking. Each component of the above setup has much higher power requirements than other normal servers I had, but much less than more than 20 of them. So this is what we mean by Green Computing … Using less power overall saves $$$$$ on outright power consumption, cooling, space requirements
b. If each server pulls 7 amps of power from of a 120 volt circuit that’s 84,000 WATTS of power required which can cost about $6000 a month (assuming about $0.10 per Kilo Watt hour). That is a lot juice and lettuce… But if you consolidate like I talked about earlier, you can run all 100 servers as virtual instances on just 6, more powerful, physical servers that draw about twice the power; we are talking 10,080 WATTS. Power consumption is cut to 12.5% (so is cost) of the previous requirements and the cost drops to about $800. (To get a real full understanding of how you pay for power click here)
c. If you are renting space at a datacenter then you have to pay for it … 100 servers means about (4) 42U racks … that takes up about 16 square feet … the same 6 servers will fit into just 1 rack now or 4 square feet … for a cost comparison Facebook rented a 300,000 square foot datacenter at a cost of about $50,000,000. The per-square-foot cost is about $166.67. Simple math here 16*166.67 = $2666.72 vs. $666.68
2. You only don’t have to manage each instance of a virtualized server as though it was physical … you just have to manage the actual physical hardware. This lowers the TCO (Total Cost of Ownership); which includes manpower needed, electrical power required, physical space required, and complexity of the physical layout. To illustrate what we are talking about think about these things
a. It might take 4 staff members to manage a computer room with 100 physical servers. If you only have 6 physical servers you can lower the man power considerably which lowers payroll costs. Let’s assume it’s a 9 to 5 shop and no weekends and each staff member makes $20.00 an hour … that means 4 guys * 40 hours a week * $20 = $3200 a week. It is not inconceivable that 1 person can manage all 6 virtualized servers, but you will have to pay him $30-$40 an hour… call it $35 an hour; that breaks it down to $1,400 a week.
So let’s do some math …. In part 1a we said that we were saving about $1,800 a week or $7,200 a month. In part 1b we were also saving $7,200 a month. In part 2a we were saving about $2,000 a month … add all that up and you are saving about $16,400 a month. How many months are there in a year? You can save this much in a year.
There are a lot more advantages of virtualization that I could get into, but upper management really doesn’t care how cool the different functions and features are, they may be mildly interested in the basic concepts of how to the systems failover and how easy it is to recover from a disaster, but they will be all ears when you tell them you can save this much money in a year … If you have to invest $75,000 dollars to realize the savings talked about here, you can figure out how fast you get a real Return On Investment.
Till next time here is a geek’s love affair with Virtualization
By Stephen Sheel
Program Director, School of Technology
Senior IT Manager