speeding up schema loads

[ghosttie]

I spend a lot of time waiting for schema loads, so I would like to speed them up.

Getting a faster multi-core CPU hasn't helped much - the CPU usage doesn't spend much time at 100%, and since the schema load is single threaded it makes no use of the rest of the cores.

Increasing the amount of memory Jade is allowed to use also didn't seem to help much.

I get the impression therefore that the process is IO bound - has anyone tried a schema load on an SSD drive or RAM drive? Is the increase is speed worth the increased cost?

[BeeJay]

A couple of tips I have for speeding up schema loads:

1. For large schema loads, try to avoid thin client and multi-user schema loads as much as possible. Single user jadload/jadloadb schema loads are considerably faster.

2. If you're happy to restore to a backup in the event of a failure, a load style of OnlyStructuralVersioning is considerably faster than LatestSchemaVersion. Even better, if you're making no structural changes to persistent classes use Current Schema Version to avoid any need for a reorg and transition phase.

I concur with your statement that the schema load is I/O bound based on the number of transaction journals that are switched during a schema load and the timing of my test deploys which take around 2 to 3 hours on my laptop, 1 hour on our standard desktop, and less than 30 minutes on our development server which has the slowest CPUs but by far the highest spec I/O subsystem.

Cheers,
BeeJay.