From: | Rosser Schwarz <rosser(dot)schwarz(at)gmail(dot)com> |
---|---|
To: | Evgeny Shishkin <itparanoia(at)gmail(dot)com> |
Cc: | Craig Ringer <craig(at)2ndquadrant(dot)com>, Niels Kristian Schjødt <nielskristian(at)autouncle(dot)com>, Jeff Janes <jeff(dot)janes(at)gmail(dot)com>, "pgsql-performance(at)postgresql(dot)org" <pgsql-performance(at)postgresql(dot)org> |
Subject: | Re: Do I have a hardware or a software problem? |
Date: | 2012-12-12 01:47:28 |
Message-ID: | CAFnxYwjVjR3zfWSHraiha_SGBLsZbEDeVS3rzZb8wV_-x3zLUg@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On Tue, Dec 11, 2012 at 5:17 PM, Evgeny Shishkin <itparanoia(at)gmail(dot)com>wrote:
> Actually most of low-end SSDs don't do write caching, they do not have
> enough ram for that.
>
AIUI, *all* SSDs do write-caching of a sort: writes are actually flushed to
the NAND media by erasing, and then overwriting the erased space, and
erasing is done in fixed-size blocks, usually much larger than a
filesystem's pages. The drive's controller accumulates writes in an
on-board cache until it has an "erase block"'s worth of them, which are
then flushed. From casual searching, a common erase block size is 256
kbytes, while filesystem-level pages are usually 4k.
Most low-end (and even many mid-range) SSDs, including Sandforce-based
drives, don't offer any form of protection (e.g., supercaps, as featured on
the Intel 320 and 710-series drives) for the data in that write cache,
however, which may be what you're thinking of. I wouldn't let one of those
anywhere near one of my servers, unless it was a completely disposable,
load-balanced slave, and probably not even then.
rls
--
:wq
From | Date | Subject | |
---|---|---|---|
Next Message | Craig Ringer | 2012-12-12 02:02:00 | Re: Do I have a hardware or a software problem? |
Previous Message | Evgeny Shishkin | 2012-12-12 01:44:24 | Re: Do I have a hardware or a software problem? |