Bacula-users

Re: [Bacula-users] Bacula spool on SSD -- solid state drive?performance testing?

2009-07-27 09:47:15
Subject: Re: [Bacula-users] Bacula spool on SSD -- solid state drive?performance testing?
From: Alan Brown <ajb2 AT mssl.ucl.ac DOT uk>
To: Kevin Keane <subscription AT kkeane DOT com>
Date: Mon, 27 Jul 2009 14:40:48 +0100 (BST)
On Fri, 24 Jul 2009, Kevin Keane wrote:

> I think even the cheapest flash memory has an MTBF of 10,000 write
> cycles, which means that an SSD used for spooling should last for 1000
> full backups - even more differential or incremental ones.

Wear levelling algorithms mean that deleted sectors aren't immediately
reused unless there are no other sectors available.

> Or is my reasoning wrong here? Without practical experience, that's
> entirely possible.

Tom's Hardware has run some pretty good overviews of SSD technology.

Until very recently I wouldn't have considered it for anything where speed
was needed.

Now that SSD drives are comparably priced to similar spec SAS drives the
equation is changing rapidly, particularly at the enterprise SLC end where
every new SSD release is significantly improved on the previous one.

As far as longevity goes, I've changed 8 spool drives out in the last 5
years, all with bad sectors or other problems indicative of the actutator
arm pivot bearing being worn out. It's annoying and time consuming, but
it's not a big deal. Changing out a SSD won't be any more difficult.

Everything else being equal, I suspect SSDs will last longer in this kind
of service - or any other kind of service where head thrash is an issue.

AB



------------------------------------------------------------------------------
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>