up vote 2 down vote favorite
share [g+] share [fb]

This involves:

  • RAID 10

  • Windows Server 2008 R2

  • Two logical drive partitions

The server in question has four drives in a RAID 10 configuration, as well as two in RAID 0. The RAID 0 is used for the OS and is behaving itself. The RAID 10 is divided into two logical drives, D and L. The L drive performs as expected, but we're seeing very poor performance on D, despite it being part of the same RAID.

How do I go about diagnosing the problem, and hopefully fixing it?

EDIT: As it turns out, I was misinformed. The L drive is partitioned off the wrong array. This solves the peculiarity of the problem, leaving a separate array slowness issue that can be diagnosed through conventional means.

link|improve this question
feedback

2 Answers

up vote 0 down vote accepted

Use perfmon to take a look at the counters for the underlying physical disk. Most interesting values are idle time, length of write queue, length of read queue and the service time (avg seconds / transfer). This Technet blog article is giving some rough orientation on how to interpret the perfmon disk counter values and find bottlenecks.

link|improve this answer
Used a different tool to get it, but this is actually the data that sent me here. sec/transfer averages .09 vs .001 on the D and L drives, respectively. – Merennulli Jun 7 '11 at 23:37
The "logical disk" values are mostly meaningless as they are just an artificial separation of the physical disk parameters. Could you post a performance graph or at least values for the given four params for your physical disk? – syneticon-dj Jun 9 '11 at 9:29
Getting the data together to reply to this showed I was misinformed about the drive configuration. Thanks for helping me get to this point. – Merennulli Jun 9 '11 at 16:38
feedback
up vote 0 down vote

You might also check to ensure the "D" volume is healthy. Checking in on its fragmentation may not hurt as well..

link|improve this answer
I checked this earlier, and it has 4% fragmentation and comes up clean with every test I know including checkdisk. – Merennulli Jun 7 '11 at 22:41
Any significant difference in use or data being stored between D & L? – user48838 Jun 7 '11 at 22:51
Yes, D has about 4x the reads and 2x the writes as L. The relevant metric, though, is that the average for reads is over 90x slower on D. This is a recent change as these two used to be nearly identical. – Merennulli Jun 7 '11 at 23:41
The read/write differences are also compounded with the RAID "hits" as well. Any way to "quieting" down D for a "sanity" check? – user48838 Jun 8 '11 at 0:05
Took a while to schedule the downtime for it, but I'll be testing this tonight. – Merennulli Jun 9 '11 at 16:26
feedback

Your Answer

 
or
required, but never shown
discard

Not the answer you're looking for? Browse other questions tagged or ask your own question.