Tuesday Nov 06, 2007
Tuesday Nov 06, 2007
Still lots of confusion about why we should virtualise and how to do it right. Just to be sure, I am talking about server virtualisation, not storage virtualisation. That I may cover later.
Virtualisation for the wrong reasons:
Easiest way to describe this, it is fashionable so we buy the latest stuff. Say we need a pair of trousers with 36inch waist, but 32inch slim fit trousers are fashionable, so we buy them. We consolidate ourselves into a bad fitting pair of trousers. Works OK for a while, however when we demand more from the trousers and sit down. The seams split open. Trouser investment wasted, embarassment we cannot hide, need to spend more time rightsizing and quickly get changed and put the correctly fitting pair of trousers on. Basically blew a lot of money for no gain, maybe satisfied our vanity for a short time.
Quick list of what too look our for if you are doing it wrong, if you or your staff are doing this, beware.
You have no historic capacity planning metrics or empiracal evidence, including peak usage periods that can show or prove that virtualisation/consolidation will make money or save money.
Other people are doing it so it must be right.
Techies are bored and need something new to do.
A virtualisation project has to be on my CV.
IT department has no answers to the companies problems, and upper managements questions so we need a new idea quick.
The solution is server virtualisation, but we do not know what the problem is.
Computer software costs are hidden from the IT budget.
Virtualisation for the right reasons:
Easiest way I can describe this, we have too many correctly fitting trousers in the cupboard, some are getting old and we are not using them (utlisation is low). The old trousers/style are an embarassment to your wife and you are not allowed to show yourself in public with them. So you take your old rarely used trousers throw them out, make space in your cupboard and instead of using them you use the newer, less worn trousers that do the job just as well.
Quick list of what to do, if you staff are doing this leave them alone.
We have capacity planning tools which can scientifically and empirically prove that virtualisation will help the company grow and become more profitable, IT will give the company an competitive edge or reduce IT costs.
I have bought too many small servers and they are not being used.
Too many systems to manage.
Not enough space in the datacenter.
Peak utlisation of a large number of the servers is very low.
I have a power and cooling problem.
Now as discussed earlier, virtualisation has been available for a very long time you do not have to buy software to do this. Unix systems like Solaris can safely run many application on one server or Operating System image, it has a 20yr track record of doing so. Say you have app A on OS A, app B on OS B, and app C on OS C. Now you can run them all on one server e.g. app A, B and C on OS Solaris. Crazy, but nothing new.
Sun has been doing virtualisation, consolidation and running many apps on one server, in a highly secure, highly available manner for 20yrs. It just happened that tight trousers came back into fashion. Sun has the whole range of well fitting and tight fitting trousers too, but only use them if you can get into them. I prefer an elasticated waste, expands and contracts automatically with capacity demands.
The cost side of the equation is extremely important, and a by the number approach is essential.
There are additional drivers on the revenue side of the equation in the area of increased IT agility, and lowering cost and operational barriers within IT to deploying new services, giving the business strategic advantages.
However "Red Shift", IT as a weapon, approaches will typically involve changes to the way IT interacts with the business. We need to stop looking at out feet.
Posted by Louis F. Springer on November 06, 2007 at 05:04 PM CET #
I think I agree, if you are saying. We must empirically calculate this, if it helps reduce systems management costs, enabling IT to adapt faster to changes in a business.
IT should be used as a business enabler rather than a cost (which is what i think you meant by looking at our feet).
Posted by Valdis on November 06, 2007 at 05:35 PM CET #
@Valdis Yes, the point is IT cost containment issues are important, but we should also be looking for drivers on the revenue side of the equation.
However, the measures on the revenue impact side may not be so easy to come up with in "real money" terms as numbers on the direct IT cost side of the equation.
For example, how much is the ability to deploy a new application in a day versus a month worth to a business? Calculating this isn't quite as straightforward as calculating power costs.
How much is a reduction in the cost or procedural barriers to deploying *new applications* worth? This depends on how valuable these potential, new applications are to the business, and not as easy to calculate as direct IT operations costs.
The fact that these "opportunity costs" are difficult to calculate does not diminish their importance. We should never hesitate to engage the business in conversations regarding these factors in strategic IT decisions.
To some degree, even having these conversations with the business is as critical as the content of the conversations, since these exchanges are more likely to draw IT closer to issues importatn to the business than discussions about IT cost expenditures on space and power.
Posted by Louis F. Springer on November 07, 2007 at 12:47 AM CET #
When launching a new product or application, which often nowadays is the same thing. The marketing depts can come up with a cost for late entry into the marketplace. For internal costs it is not too difficult to calculate the difference between 2 days work and 4 weeks work. Also, what other projects are not initiated. Good IT depts can calculate this. However, inter departmental politics plays a big part. Good to see that other people think the same way, thanks for the comments.
Posted by Valdis on November 07, 2007 at 09:49 AM CET #