Anyone who’s been in this industry for longer than a decade will know that some of what IT vendors say needs to be taken with a pinch of salt. Virtualisation holds great promise, so we are told – but yet so did blade servers, grid architectures, enterprise management solutions, application service providers... the list goes on. …
Virtualisation is taking me ...
... to the shop to buy more memory.
Professionally there is some merit in running VMs (after all, IBM has made a mainframe business out of it). Although not as much as the marketing bumf says: most of the benefits can be achieved by consolidating workloads onto a single server, without the overheads, administration and duplication of multiple copies of the same old O/S.
Where it does win is with my old mum. She thinks her XP 'pooter is running native! In fact it's running as a Virtualbox guest on top of Ubuntu 8.04. This, and VNC, does make remote administration and support (a must with m.o.m-s the world over) much easier as I can access the system underneath XP whenever necessary to reboot it etc. All without having to go though the joys of getting a 70 year-old, 150 miles away, to press the right combination of keys, buttons and wotsits in the right order while holding the phone, too.
 actually she doesn't care what it is or what it runs on - just so long as it works.
 not that age is a driver, work provides the same opportunities with 20-60 y-o's as well.
What a load of bull
"some of what IT vendors say needs to be taken with a pinch of salt" ??? Excusame?? "some of"? Let's just say that _everything_ that a vendor say needs to be taken with humongous pile of salt.
"making IT a bit more dynamic" WTF? What happened to "planning" and "management"? Instead of coming up at 8pm on friday evening with "let's put 3 new servers up" how about thinking about it a couple of week before so we can figure out if we can run that shit on the machines that we already have or we need to buy or rent something else? Bad planning on your side should never be emergency in the IT department, but of course it is. All the time.
“people are expensive, automation is good" so that's why everybody is trying to ditch peoples and relies on pile of crap hosted god-knows-where and run by untrained chimps someplace 8 hours from you?
The reality is that management doesn't know shit about how to run IT or how to run anything really, but they certainly do like to throw buzzwords around to show that they are "in the loop".
@AC I hear you
Aye, planning and management should be in there. The "more dynamic" thought could have been rephrased "a bit less klunky when it comes to actually implementing anything"... I agree that more planning would be a great thing. From other feedback I'm not sure many organisations have it quite as bad as you are suggesting, but I can remember working at an insurance firm a few years back, and sitting in that very room when the ops manager was "told" what would be required of him the week after, in terms of new kit to support. So amen to that.
Generally for development
We've found that virtualisation works really well for development kit, testing/demo sites and the like. It's saved us the cost of additional servers because we can just fire up and tear down exactly what we need, when we need it, which in a flexible development environment is great.
For anything customer-facing (ie, production servers), we still do it old-school, simply because you tend to have fewer actual apps on that end (one app might be tested on a whole range of machines, but our hosted version only needs one), the user demands on them are far higher, and you are far less likely to want to mess with it.
Virtualization is Big
It is a major tool for consolidation of servers and it is magical for thin clients, a very successful form of virtualization. Thin clients really save capital cost of equipment and power consumption because one server can run hundreds of $250 thin clients each using 30 watts or so. Where I worked last year, they could run all their services on one or two machines with virtualization. The saving in hardware and power consumption and space would have been huge. If you get improved security from virtualization it is a huge plus. We do not need more servers as much as we need more services. Virtualization does that very well.
Unix GNU/L has done this for years as a multiuser system you don't need vitualisation.
Why not for production machines?
Virtualization provides all kinds of benefits and it will - of course - depend on your specific situation. IMO on of the main benefits is making the (guest) OS independent from the hardware it runs on. The physical machine dies and cannot easily be repaired? No problem, just take a completely different server and run it there without much hassle. Things like resource scheduling, power management, high availability and fault tolerance may be important in your choice of virtual or physical. In some cases it may be necessary to combine both, YMMV.
One of the big negatives is that people think they can simply keep on adding servers left and right, creating a management and licensing nightmare.
The good, the bad, the ugly
The good: I can stand up machines on the network segment I need to quickly and easily. While you can do a lot by consolidating workloads on one OS the fact remains that sometimes you need to segregate the applications, and if you work with a lot of different network segments virtualization is a great way to deal with that.
I can juggle resources more easily, and the ability to migrate from virtual to physical or vice versa comes in handy.
I can make high availability a lot cheaper to implement and maintain for a lot more applications.
I don't have to manage HBAs on a lot of physical machines....only one. (Ok, that's a pet peeve. I hate maintaining the client side of the SAN infrastructure.)
You can mix DR with Dev and QA environments and be able to say with a straight face that DR can give you 80%+ of production as long as they let you shut down the Dev and QA stuff if you need to use it.
The Bad: Management uses virtualization to accomplish an "out of sight out of mind" mentality on resources. You need to produce a lot of really good reports about how badly you're oversubscribing before they take you seriously. Even then they prefer to not do anything until something bad happens.
The Ugly: When bad things happen, the consequences are expanded.
On "the bad"...
Can I be blunt here - I don't think we've seen the worst of it yet from a manageability perspective. The risk of IT copping out is high, particularly if faced with too many demands for new servers/services. Best practice in these areas is still being defined - which is why all feedback is so very valuable! Thanks!
The fingers point, and having pointed, move on...
Virtualisation is wonderful; now, instead of simply having the hardware vendor and software vendors shake their heads and tell you it's "the other guy's problem" you have a three-sided finger-pointing, consultancy-fattening orgy of blame; 'Its the hardware", "Its the software", "its the virtualisation layer" ... around and around until the support contract runs out.
Paris, because she knows all about the three way thing