Will a 6 disk Raid 5 array beat the 6 disk Raid 10 array on reads?see
http://techrepublic.com.com/5208-11184-0.html?forumID=55&threadID=183003
http://beau.org/pipermail/whitebox-users/2005-July/006933.html
"Dave" wrote:
> Will a 6 disk Raid 5 array beat the 6 disk Raid 10 array on reads?
>|||I still havent found the answer to this... does anyone have an online
resource or know of any benchmarks?
CitizenPips wrote:
> see
> http://techrepublic.com.com/5208-11184-0.html?forumID=55&threadID=183003
> http://beau.org/pipermail/whitebox-users/2005-July/006933.html
> "Dave" wrote:
> > Will a 6 disk Raid 5 array beat the 6 disk Raid 10 array on reads?
> >
> >|||This is a multi-part message in MIME format.
--090606080509080509000803
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Looking purely at read performance (and ignoring other factors like
write performance, cost, complexity, fault tolerance, etc.) both RAID 5
& RAID 10 offer excellent performance. Both configurations will
probably be close enough in read performance to each other that other
factors (like those mentioned above) will probably be more influential
in any decision to purchase one over the other.
However, in your specific example (6 disks in RAID 5 or 6 disks in RAID
10) I would think the RAID 5 array should be slightly (very slightly)
faster and only when not under heavy load. This is because for any
particular read against the RAID 5 set there are 5 disks (the slice on
the 6th disk would just be parity) in the set to read data from (these
reads can happen in parallel). For the RAID 10 there would be 3 disks
to read data from, although for each of those 3 disks the IO subsystem
could choose the primary or the mirror depending on current queue
lengths (which is not an option with the RAID 5 set). So you have the
possibility of 5 simultaneous reads on potentially loaded disks with
RAID 5 vs the possibility of 3 simultaneous reads on potentially less
loaded disks with RAID 10. As a rule of thumb, the more spindles you
can read from simultaneously, the quicker that read will be, but this
assumes no queuing on those spindles. When dealing with a mirror, as
RAID 10 is (combined with striping), the IO subsystem can choose from
either the primary or mirror disk to read from (if one is too
overworked). So one advantage of "more spindles" is pretty much
cancelled out by the advantage of "choosing an alternate disk".
Toss a coin as either could be faster on any given day depending on all
the other surrounding factors.
BTW, _Inside SQL Server 2000_ by Kalen Delaney has a good section (in
chapter 4, pp. 127-137) on RAID configs.
Hope this helps.
--
*mike hodgson*
http://sqlnerd.blogspot.com
Dave wrote:
>I still havent found the answer to this... does anyone have an online
>resource or know of any benchmarks?
>CitizenPips wrote:
>
>>see
>>http://techrepublic.com.com/5208-11184-0.html?forumID=55&threadID=183003
>>http://beau.org/pipermail/whitebox-users/2005-July/006933.html
>>"Dave" wrote:
>>
>>Will a 6 disk Raid 5 array beat the 6 disk Raid 10 array on reads?
>>
>>
>
>
--090606080509080509000803
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
<tt>Looking purely at read performance (and ignoring other factors like
write performance, cost, complexity, fault tolerance, etc.) both RAID 5
& RAID 10 offer excellent performance. Both configurations will
probably be close enough in read performance to each other that other
factors (like those mentioned above) will probably be more influential
in any decision to purchase one over the other.<br>
<br>
However, in your specific example (6 disks in RAID 5 or 6 disks in RAID
10) I would think the RAID 5 array should be slightly (very slightly)
faster and only when not under heavy load. This is because for any
particular read against the RAID 5 set there are 5 disks (the slice on
the 6th disk would just be parity) in the set to read data from (these
reads can happen in parallel). For the RAID 10 there would be 3 disks
to read data from, although for each of those 3 disks the IO subsystem
could choose the primary or the mirror depending on current queue
lengths (which is not an option with the RAID 5 set). So you have the
possibility of 5 simultaneous reads on potentially loaded disks with
RAID 5 vs the possibility of 3 simultaneous reads on potentially less
loaded disks with RAID 10. As a rule of thumb, the more spindles you
can read from simultaneously, the quicker that read will be, but this
assumes no queuing on those spindles. When dealing with a mirror, as
RAID 10 is (combined with striping), the IO subsystem can choose from
either the primary or mirror disk to read from (if one is too
overworked). So one advantage of "more spindles" is pretty much
cancelled out by the advantage of "choosing an alternate disk".<br>
<br>
Toss a coin as either could be faster on any given day depending on all
the other surrounding factors.<br>
<br>
BTW, <u>Inside SQL Server 2000</u> by Kalen Delaney has a good section
(in chapter 4, pp. 127-137) on RAID configs.<br>
<br>
Hope this helps.<br>
</tt>
<div class="moz-signature">
<title></title>
<meta http-equiv="Content-Type" content="text/html; ">
<p><span lang="en-au"><font face="Tahoma" size="2">--<br>
</font></span> <b><span lang="en-au"><font face="Tahoma" size="2">mike
hodgson</font></span></b><span lang="en-au"><br>
<font face="Tahoma" size="2"><a href="http://links.10026.com/?link=http://sqlnerd.blogspot.com</a></font></span>">http://sqlnerd.blogspot.com">http://sqlnerd.blogspot.com</a></font></span>
</p>
</div>
<br>
<br>
Dave wrote:
<blockquote
cite="mid1144872987.515028.64430@.u72g2000cwu.googlegroups.com"
type="cite">
<pre wrap="">I still havent found the answer to this... does anyone have an online
resource or know of any benchmarks?
CitizenPips wrote:
</pre>
<blockquote type="cite">
<pre wrap="">see
<a class="moz-txt-link-freetext" href="http://links.10026.com/?link=http://techrepublic.com.com/5208-11184-0.html?forumID=55&threadID=183003</a>">http://techrepublic.com.com/5208-11184-0.html?forumID=55&threadID=183003">http://techrepublic.com.com/5208-11184-0.html?forumID=55&threadID=183003</a>
<a class="moz-txt-link-freetext" href="http://links.10026.com/?link=http://beau.org/pipermail/whitebox-users/2005-July/006933.html</a>">http://beau.org/pipermail/whitebox-users/2005-July/006933.html">http://beau.org/pipermail/whitebox-users/2005-July/006933.html</a>
"Dave" wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Will a 6 disk Raid 5 array beat the 6 disk Raid 10 array on reads?
</pre>
</blockquote>
</blockquote>
<pre wrap=""><!-->
</pre>
</blockquote>
</body>
</html>
--090606080509080509000803--|||Thanks Mike! That was an excellent explanation! I am going to request
a RAID 5 Array to be attached to my Dell 2850 so I can benchmark one
Array vs the other.
I will test large SELECT/JOINS, inserts, a couple intensive stored
procedures that perform both read/writes, and run Quest's benchmark
software (free version).sql
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment