Re: How is random_page_cost=4 ok?

Поиск
Список
Период
Сортировка
От Michael Renner
Тема Re: How is random_page_cost=4 ok?
Дата
Msg-id 48EF4E24.4050708@amd.co.at
обсуждение исходный текст
Ответ на How is random_page_cost=4 ok?  (Gregory Stark <stark@enterprisedb.com>)
Ответы Re: How is random_page_cost=4 ok?  (Gregory Stark <stark@enterprisedb.com>)
Список pgsql-hackers
Gregory Stark schrieb:

> Te reason I'm wondering about this is it seems out of line with raw i/o
> numbers. Typical values for consumer drives are about a sustained throughput
> of 60MB/s ( Ie .2ms per 8k) and seek latency of 4ms. That gives a ratio of 20.
> 
> Server-class drives have even a ratio since a 15kRPM drive can have a
> sustained bandwidth of 110-170 MB/s (48us-75us) and an average seek latency of
> 2ms giving a ratio of 27-42. And of course that doesn't include the effects of
> a RAID array which magnifies that ratio.

Hi Gregory,

I think your numbers are a bit off:

For "Consumer drives" (7.200 RPM SATA 3.5"), seek times are much worse, 
in the area of 8-9ms (see [1]), but sustained sequential read numbers 
are noticeable higher, around 80-90MB/sec.

For "Server Drives" 3-4ms are more realistic ([2], [3]) for average 
seeks and the 110-170MB/sec are highly exaggerated.


Unfortunately I have only 2.5" SAS 10k drives and no FreeBSD here, 
otherwise I could provide some real world numbers; the diskinfo tool in 
[3] looks really nice (and makes me crave FreeBSD).

best regards,
Michael

[1] http://h18004.www1.hp.com/products/quickspecs/13021_div/13021_div.html

[2] http://h18004.www1.hp.com/products/quickspecs/12244_div/12244_div.html

[3] 
http://blog.insidesystems.net/articles/2007/04/09/unscientific-15k-v-10k-sas-drive-comparison


В списке pgsql-hackers по дате отправления:

Предыдущее
От: Andrew Sullivan
Дата:
Сообщение: Re: Updates of SE-PostgreSQL 8.4devel patches
Следующее
От: Matthew Wakeling
Дата:
Сообщение: Re: CREATE DATABASE vs delayed table unlink