From: | David Gilbert <dgilbert(at)velocet(dot)ca> |
---|---|
To: | nikolaus(at)dilger(dot)cc |
Cc: | dgilbert(at)velocet(dot)ca, chris(at)ruprecht(dot)org, mallah(at)trade-india(dot)com, pgsql-admin(at)postgresql(dot)org |
Subject: | Re: H/W RAID 5 on slower disks versus no raid on faster HDDs |
Date: | 2002-11-26 02:42:54 |
Message-ID: | 15842.57262.323171.410126@canoe.velocet.net |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-admin |
>>>>> "nikolaus" == nikolaus <nikolaus(at)dilger(dot)cc> writes:
nikolaus> Of course can software RAID be faster than hardware RAID.
nikolaus> But then you are not comparing the best offerings of each
nikolaus> category. Software RAID is usually cheaper than hardware.
nikolaus> But again you may be able to construct a product pairing
nikolaus> where it is reverse.
No... I admit... if you're willing to spend huge $$'s, there exist
hardware solutions that outpace software. It usually means that you
have hardware that is more powerful than your main system, though.
Often more costly, too.
nikolaus> raw versus file system is an old battle. raw is faster
nikolaus> because it does not have the filesystem overhead. However,
nikolaus> raw is a little harder to administer. Question is how much
nikolaus> faster raw is and if it is worth the trouble. Numbers
nikolaus> usually quoted are 10-20%. And here things may be changing
nikolaus> as operating systems and chaching algorithems get better to
nikolaus> mask the file system overhead. As I mentioned the question
nikolaus> is mute for PostgeSQL because raw is currently not
nikolaus> supported.
Well... I don't think your raw write performance (in, say, FreeBSD)
would be any more than 5% better than filesystem writing. Old
filesystem overhead consisted of the in-order writes that were
required to update filesystem metadata... which (consequently) was
about 20% in-order (vs. un-ordered) writes.
Softupdates (FreeBSD) only requires 1%-ish (or less) in-order writes
(or to put it another way: 1% of writes chosen before they would
otherwise happen).
DAve.
--
============================================================================
|David Gilbert, Velocet Communications. | Two things can only be |
|Mail: dgilbert(at)velocet(dot)net | equal if and only if they |
|http://daveg.ca | are precisely opposite. |
=========================================================GLO================
From | Date | Subject | |
---|---|---|---|
Next Message | Gaetano Mendola | 2002-11-26 06:52:18 | Child itemid in update-chain marked as unused - can't continue repair_frag |
Previous Message | David Gilbert | 2002-11-26 02:36:30 | Re: H/W RAID 5 on slower disks versus no raid on faster HDDs |