Tuesday, April 25, 2006

Scale Out Panel

I am at the scale out panel where DBA gurus from Technorati, MySQL (Brian Aker), Google and Yahoo! are answering questions about scaling.

Some Windows based sites are storing up to 30-40 TB of data using MyISAM tables. That site is masochist.com (joke)

A question was asked whether the date in database should be stored as formatted or stored as time only and then formatted by application. The answer given was (the room was jam packed so I couldn't see the faces) that optimization like this should be done at the end. At first, you should try to add indexes where ever needed. Once all the indexes have been added, only then we should move with other smaller optimizations.

Power consumption having effect on business as you scale out?
There is power, power per watt etc. Power is huge concern for us (Technorati).

Some CPUs use a lot less power than others. At Yahoo! and Google, power is a major concern. We have to find machine that use less power (may be slower disk seeks) and slower CPUs that use less power. Power consumption is a huge issue but only for very large organizations (Yahoo!).

When we went to add some more servers, we weren't being given a data strip because they were out and we have run into such problems.

Load balancing and RAIDs based implementations are easier to scale out.

64bit gives more memory to InnoDB with less power consumption. Use it if at all possible.

Update: Thank you Brian for pointing out that it was TB, not GB. It was a typo on my part that I have corrected.

2 comments:

Anonymous said...

Hi!

It was terabytes, not gigabytes. Lots of users have 30-40 gigs of data in MyISAM.

Cheers,
-Brian

Frankly Speaking! said...

Thank you Brian for pointing out that it was TB, not GB. It was a typo on my part that I have corrected.

Cheers,
Frank (Farhan)