Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


LSI MegaRAID sucks?! - Page 2
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

LSI MegaRAID sucks?!

2»

Comments

  • Yeah, in my tests it was on a brand new box as well (no previous writes), therefore my conclusion LSI sucks.

  • letboxletbox Member, Patron Provider

    Ok, i do some configured.

    here is the test

    [root@xxxxxx]# ./ioping . -c 10

    4096 bytes from . (ext4 /dev/sda2): request=1 time=0.1 ms
    4096 bytes from . (ext4 /dev/sda2): request=2 time=0.1 ms
    4096 bytes from . (ext4 /dev/sda2): request=3 time=0.1 ms
    4096 bytes from . (ext4 /dev/sda2): request=4 time=0.6 ms
    4096 bytes from . (ext4 /dev/sda2): request=5 time=0.6 ms
    4096 bytes from . (ext4 /dev/sda2): request=6 time=0.7 ms
    4096 bytes from . (ext4 /dev/sda2): request=7 time=0.9 ms
    4096 bytes from . (ext4 /dev/sda2): request=8 time=0.6 ms
    4096 bytes from . (ext4 /dev/sda2): request=9 time=0.1 ms
    4096 bytes from . (ext4 /dev/sda2): request=10 time=0.6 ms

    --- . (ext4 /dev/sda2) ioping statistics ---
    10 requests completed in 9006.2 ms, 2239 iops, 8.7 mb/s
    min/avg/max/mdev = 0.1/0.4/0.9/0.3 ms

    [root@xxxxx]# ./ioping . -R

    --- . (ext4 /dev/sda2) ioping statistics ---
    23109 requests completed in 3000.1 ms, 13720 iops, 53.6 mb/s
    min/avg/max/mdev = 0.0/0.1/5.9/0.0 ms

    [root@xxxxx]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync

    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 2.51498 s, 427 MB/s

    [root@xxxxx]# dd if=/dev/zero of=test bs=16k count=16k conv=fdatasync

    16384+0 records in
    16384+0 records out
    268435456 bytes (268 MB) copied, 0.327438 s, 820 MB/s

    [root@xxxxxx]# top -c

    top - 08:31:15 up 15 min, 2 users, load average: 0.01, 0.10, 0.09
    Tasks: 276 total, 1 running, 275 sleeping, 0 stopped, 0 zombie
    Cpu(s): 0.0%us, 0.0%sy, 0.0%ni,100.0%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st

    on busy SAS Raid 10

    [root@xxxx]# ./ioping . -R

    --- . (ext3 /dev/root) ioping statistics ---
    2793 requests completed in 3000.7 ms, 10657 iops, 41.6 mb/s
    min/avg/max/mdev = 0.0/0.1/30.6/0.7 ms
    [root@xxxxxx]#

  • You will get lower latency and higher IOPS if you use the software SSD caching (the flashcache module).

  • @Nick_A said: Coincidentally, if anyone wants to buy about 8 128GB 840 Pros, I know someone with extras...

    I may be interested. PM me your price.

  • Nick_ANick_A Member, Top Host, Host Rep

    @Brandon said: I may be interested. PM me your price.

    Alright, let me figure out if I want to try them on the 9266 first.

  • Nick_ANick_A Member, Top Host, Host Rep

    @concerto49 said: Not at the moment, but it is very interesting in that a dd test on / gets almost double the IO compared to a dd test on /vz. That's the only "issue" I've found so far on it, which is strange.

    Are you using LVM or anything?

  • @Nick_A said: Are you using LVM or anything?

    LVM. Doesn't happen on Intel 520 though - definitely something Samsung 840 Pro specific.

  • Nick_ANick_A Member, Top Host, Host Rep

    Poop

  • We've switched our new build to intel 520s. It's not ideal; however the performance difference is negligible and both offer 5 year warranties

  • @ShardHost said: We've switched our new build to intel 520s. It's not ideal; however the performance difference is negligible and both offer 5 year warranties

    What do you mean the performance difference is negligible? Are you saying the Intel 520 and Samsung 840 Pro are about equal if hung of a LSI card? I was considering ordering a few 520's to compare, but if you've already done so...

  • @miTgiB said: What do you mean the performance difference is negligible?

    Forgetting about RAID and LSI issues, the Samsung 840 Pro is miles ahead of an Intel 520. It's not negligible for sure.

  • @concerto49 said: Forgetting about RAID and LSI issues

    Unfortunately, this is an issue that cannot bet forgotten about, as it seems there is an issue with the interaction of the 9271 raid card and Samsung 840 SSD, so the firmware of one or both needs something tweaked it seems, but in the mean time I'd rather find an alternate solution

  • @miTgiB said: so the firmware of one or both needs something tweaked it seems, but in the mean time I'd rather find an alternate solution

    It's the LSI card. It's currently not on the list of certified SSDs for cachecade. They "expect" it to be, so possibly a patch / update on their end.

    I would stick with Samsung 830 or Intel 520 for now. Prefer Samsung 830 but anyhow.

  • @concerto49 said: Prefer Samsung 830 but anyhow.

    I do have many 830's in service, I guess it is just a matter of hunting them down for now as NewEgg and SuperBiiz seem to have been out of stock on the 830's for quite some time now. Sometimes simple availability dictates what I buy.

  • @miTgiB said: I guess it is just a matter of hunting them down for now as NewEgg and SuperBiiz seem to have been out of stock on the 830's for quite some time now. Sometimes simple availability dictates what I buy.

    Possibly phasing them out for Samsung 840, but this non-PRO uses TLC instead of MLC and wouldn't recommend it on a server at all.

  • Nick_ANick_A Member, Top Host, Host Rep

    @miTgiB said: I do have many 830's in service, I guess it is just a matter of hunting them down for now as NewEgg and SuperBiiz seem to have been out of stock on the 830's for quite some time now. Sometimes simple availability dictates what I buy.

    If you need 128GB 830s, PM me. I don't want to publish my current source for fear that they'll run out soon D:

  • letboxletbox Member, Patron Provider

    @Nick_A said: If you need 128GB 830s, PM me. I don't want to publish my current source for fear that they'll run out soon D:

    I have alot of 128GB 830s too -:P

Sign In or Register to comment.