Interesting numbers on a CREATE INDEX

От: Jim C. Nasby
Тема: Interesting numbers on a CREATE INDEX
Дата: ,
Msg-id: 20050423030920.GZ58835@decibel.org
(см: обсуждение, исходный текст)
Список: pgsql-performance

Building a single-column index on a dual opteron with 4G of memory, data
on a 4 SATA RAID10; OS, logs and tempsace on a SATA mirror, with
sort_mem set to 2.5G, create index is actually CPU bound for large
portions of time. The postgresql process and system time are accounting
for an entire CPU, and systat (this is a FreeBSD5.2 box) is generally
showing 80% utilization on the RAID10 and 40% on the mirror.

Not a performance problem, but I thought some people might be
interested. The RAID10 is doing about 28-32MB/s, I would think this
wouldn't be enough to swamp the CPU but I guess I would be thinking
wrong.

BTW, the column I'm indexing is a bigint with a low correlation.
--
Jim C. Nasby, Database Consultant               
Give your computer some brain candy! www.distributed.net Team #1828

Windows: "Where do you want to go today?"
Linux: "Where do you want to go tomorrow?"
FreeBSD: "Are you guys coming, or what?"


В списке pgsql-performance по дате сообщения:

От: Greg Stark
Дата:
Сообщение: Re: Bad n_distinct estimation; hacks suggested?
От: "Joel Fradkin"
Дата:
Сообщение: Re: Joel's Performance Issues WAS : Opteron vs Xeon