Re: NetBSD on a database server - WAPBL, softdeps

[Mindaugas Rasiukevicius <rmind%netbsd.org@localhost>]

Thor Lancelot Simon <tls%panix.com@localhost> wrote:
> My point is that any journaling by the filesystem, with the consequent
> duplication of I/O, is going to seriously degrade database runtime
> performance in return for a very small boot-time benefit after a system
> crash.

OK, we should probably have started here.  I think it depends on database
workload, e.g. I/O should be minimised with heavy caching in memory for
typical OLTP-type and datawarehouse-type (except read is usual concern here)
databases.  Agree that I/O and thus journaling may have an effect with
intensive write (insert/update) streams.

My point is that "do not use journaling" is not a universal solution, for
many workloads it has small or even no effect to performance.

> Anyone who cares _that_ much about recovery time after a crash should be
> duplicating the database -- you can't be sure the system will recover *at
> all* after a crash.

Of course, but available resources get in question here.  One can suggest to
use clustering for write-intensive databases, but that may not be possible.

> If it's that much of an issue, use raw partitions.
> 

Can do with Oracle, but not with PostgreSQL (yet?). :)

-- 
Mindaugas