please do not place your comments to more than one blog post.
As I mentioned you should really try to benchmark on your workload rather than use existing benchmarks which might not be relevant for your application.

Regarding framework to setup benchmarks – perhaps Quest Benchmark Factory could do

Speaking about MySQL 4 -> 5 upgrade there are variants.

1) If you application was not well optimized for MySQL 4 it may work much faster on MySQL 5 – for example if you used OR on different columns which not works well with index merge.

2) You may get benefit from MySQL optimizations – like index merge may improve performance, some changes making group by more efficient, improved join optimizer and all these little things – it is close to one but more subtle.
Also you could get improvement using Innodb tables due to Compact storage format, especially if your database is close to memory size – in this case saving 10-20% space is very important.

3) The way to benefit from MySQL 5.0 is to use its features but of course it should make sense for your application – for example as you could see in benchmarks DBT2 benchmarks can be made 30% faster with 5.0… but if you use Stored procedures. Also you can use triggers instead of separate updates and these kind of things.

Would love to hear your thoughts on how after upgrading to MySQL 5, an application can achieve the same performance as when it was on MySQL 4 since I’ve been reading that upgrading (without making significant changes) can reduce performance.

Frank

There is a common perception that open source code is leaner and meaner by nature than closed source. For example, compare Oracle and MySQL. No one can doubt that Oracle’s database offers more bells and whistles, which require a lot……