Talking About MySQL

While I am talking about database and scaling, Peter Zaitsev has some notes on new Innodb features being discussed at the recent MySQL Users Conference.

Notably he mentions:

Many Innodb IO settings were optimized on systems with 1-2 SATA drives so they are far from optimal for server grade IO subsystems.

Which would further reinforce the notion that creating lots of little piles of data is a much better way to scale than creating a few large piles.

In fact it is better on a number of fronts:

• Having lots of machine increases the amount of memory you can bring to bear on a system, better to have 10 machines with 8GB each than one machine with 64GB.
• Having lots of machines increases the aggregate bandwidth available to you across the board. Using the precious example, you have 10 busses working for you rather than one.
• Having lots of machines increases your redundancy. If one out of 10 is broken, you have 10% degradation, whereas one out of one is 100% degradation.
• And so on and so forth… You get the picture…