This article is more than 1 year old

Excessively fat virtual worlds – come on, it's your guilty secret

Does my server look big in this?

Where's the catch?

Initially, you may find they will be very skeptical about claims of “same performance for less money”. My advice is to start with the lowest-hanging fruit, the more over-specified the better. Put the savings in black and white and agree to test the new set-up. Be sure to tell them that if it doesn’t work out, they can go straight back to the old configuration.

Once you have agreed a time to switch to a less extreme configuration, change the size and let it run as usual. After a week, review performance with the application owner – the reporting tools should indicate if the reduction in size has been effective.

Discuss any negative effects found and deal with them, as angry customers won’t let you do it again if it isn’t running as well as it did previously. After the business sees that right-sizing a virtual machine is not only good practice, but also saves money without impacting services, they will want the rest done in double-quick time.

There are two caveats, however. Firstly, Java-based applications tend to not react very well if they go short on RAM. Exercise caution where Java is concerned. Secondly, it is not always appropriate to reduce resources below an application's minimum requirements, as this can cause application failure in unpredictable ways.

All of the above is useful, but there is one left thing to address: before a machine gets virtualised or built from scratch, review what is being requested. P2V (Physical to Virtual) conversions have the advantage that the performance profile is already known and makes right-sizing it in the virtual machine world slightly easier.

New builds do tend to have a suck-it-and-see nature, however, but any sizable company should have a post-install process to review utilisation in production and make changes as needed. The big sell is the fact that most good hypervisor platforms will allow hot add of RAM and CPU, so they can be ratcheted up without downtime. Taking away capacity however, will require downtime.

Any normal-sized virtual machine estate could benefit from such treatment, even when taking into account the cost of software to evaluate the estate. It not only frees up previously purchased capacity for new machines but also ensures that the cost versus performance profile is maintained. ®

More about

TIP US OFF

Send us news


Other stories you might like