> MySQL does not scale well beyond 4 processors, at least not like PostgreSql
> does. MySQL seems to hit a plateau rather quickly. If XtraDb's modified
> Innodb plugin scales better, then fine. But I haven't seen any benchmarks
> showing the speed improvements relative to the number of processors used and
> is something I'd really like to see.
You can find such benchmarks on our blog. And Mark Callaghan and
maybe some others have benchmarked it too. Of course, we would love
to see more independent benchmarks. Vadim considers that we've solved
scalability problems in XtraDB up to 16 cores, and I agree, though I
am less of an expert than he is. However, many problems in MySQL
itself remain even if all the storage engines are fixed.
>> As others said, the major bottlenecks are likely to be internal (to the
>> locking and disk access speed.
> Of course. When it comes to MySQL, I would invest more money into more
> memory and fast SSD drives rather than more CPU's. You'll get a bigger bang
> for the buck. :)
None of MySQL's current storage engines takes advantage of a lot of
memory or fast SSD drives either, in my opinion. Not like they could,
Have you seen our (or Jignesh Shah's, or Matt Yonkovit's) benchmarks
and discussion on SSD drives? When you disable the (unsafe,
non-battery-backed) write cache, suddenly they aren't so fast anymore.