lun stats and what does it mean?

[lebedevanton]

Hi folks, I want to ask a question about lun stats -o command, and what I can take from that output to incr

ease my performance on certain luns.

We have recently started to host exchange 2007 on our filers.

we are running a fas31xx series, in a cluster.

we have 2 exchange server at this point, one is on one head, and the other is on the other.

this is the output for one of my filers:

lun stats -o                               

   /vol/hqnycexc02_db2_log2/hqnycexc02_db2_log2/hqnycexc02_db2_log2  (3 days, 4 hours, 13 minutes, 18 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes
        32302565        67607425        214273    2577591    53750      0       192         0
    /vol/hqnycexc02_db2_sg2/hqnycexc02_db2_sg2/hqnycexc02_db2_sg2_j  (3 days, 4 hours, 14 minutes, 51 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes
        64253882        64056654        7682825   1277503    53792      0       192         0
    /vol/hqnycexc02_pf_log/hqnycexc02_pf_log/hqnycexc02_pf_log_h  (3 days, 4 hours, 32 minutes, 51 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes
        2323021         4677783         26171     215452     53820      0       192         0
    /vol/hqnycexc02_pf_pfsg/hqnycexc02_pf_pfsg/hqnycexc02_pf_pfsg  (3 days, 4 hours, 35 minutes, 37 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes
        3474315         2640938         420091    57868      53844      0       192         0
    /vol/hqnycexc02_snapinfo/hqnycexc02_snapinfo/hqnycexc02_snapinfo_i  (3 days, 5 hours, 3 minutes, 42 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes
        202976          102013630       35255     209244     54676      0       192         0
    /vol/hqnycexc02_db1_log1/hqnyexc02_db1_log1/hqnyexc02_db1_log1_f  (3 days, 5 hours, 9 minutes, 54 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes
        67023237        140539854       423406    4869603    53944      0       192         0
    /vol/hqnycexc02_db1_sg1/hqnycexc02_db1_sg1/hqnycexc02_db1_sg1_e  (3 days, 6 hours, 46 minutes, 22 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes
        142375190       137890847       16900537  3174853    54094      0       192         0
    /vol/hqnycexc02_binaries/hqnycexc02_binaries/hqnycexc02_binaries_d  (4 days, 4 hours, 8 minutes, 6 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes
        767334          3921820         39500     153200     54939      0       324         0
    /vol/hqnycexc02_system/hqnycexc02_system/hqnycexc02_system_c.lun  (9 days, 3 hours, 57 minutes, 12 seconds)
        Read (kbytes)   Write (kbytes)  Read Ops  Write Ops  Other Ops  QFulls  Partner Ops Partner KBytes

by taking a look at this output, what can you suggest ?

Ive read about reallocate, and I ran it once before, to optimize large volumes and aggregates.

Anything else that I can do to optimize the system ?

Thank you!

Anton

[adamfox]

I'm not sure if you are going to gleam much from this in terms of optimization with the exception of partner ops/kbs.  Those should be at or near 0 unless you've good paths to your hosts for some reason.

Another thing that is more interesting would be to zero them out and then look at what you're getting over a smaller amount of time.  3 days isn't very useful.  The command may have a -z flag to zero out the stats.

Otherwise you'll have to do the math yourself.

But just telling you how many ops you are doing is nice, but probably not the best in terms of optimization.

[lebedevanton]

so youre saying I should do a lun stats -z and then do a -o to see in a short amount of time what the out put would be ?

[adamfox]

I'm saying that whatever you hope to glean from those particular stats are usually more meaningful over shorter time periods. This is especially true if you make some change. The effects of that change or lack thereof are more evident that way.

However, I'm still not sure what you are looking for here.

-- Adam Fox

[lebedevanton]

I want to see that I am getting the most out of my storage, and I have good I/O on my exchange.

I know thats a very broad statement, but perhaps if you can point me to the right starting point, i would appreciate it.  Thanks again.

[adamfox]

Hmm...that's tough when you don't necessarily have a problem or a good basis of comparison.

My best recommendation would be to get a hold of the best practices guide for Exchange 07 on NetApp and ensure that you are following as much of that as possible in your environment. It should be on the external site in the Library. Or your SE should be able to get one for you.

But just looking at lun stats other than partner ops may not tell you much as your the raw ops will depend on the load, etc.

-- Adam Fox

[lwei]

You can look at sysstat to get a broad view of how the storage system is doing, cpu, disk, IOPS, etc. Thanks,   -Wei