Visit Open-E website
Page 1 of 2 12 LastLast
Results 1 to 10 of 12

Thread: Hardware Recommendations?

  1. #1
    Join Date
    Apr 2009
    Posts
    62

    Default Hardware Recommendations?

    Hey Guys,

    I see a lot of people post their specs on these forums and a lot of good performance numbers coming from some of them.

    I'm interested in building a homegrown machine, preferably 12+ hot swap SAS drive bays.

    Anyone have any hardware recommendations? (ie. RAID Controllers, Mainboards, Drives, Chassis) I'm looking for good DSS hardware support, not really interested in experimental hardware.

    I will be using Intel 10GbE NICs to connect this new Storage to my existing storage network.

    If you have benchmarks to back up the numbers that would be great.

    Thanks for your input!

    Drew

  2. #2

    Default

    Hi,

    we have build our system in april-may.
    Hardware is redundant:
    Mainboard Supermicro X8DT3
    Chasis SC836
    CPU 1xIntel Xeon 5550W
    RAM 24GB (6x4GB)
    Raid Controller Adaptec ICP Vortex 8165BR with BB 256 Cache
    HDD 16x Western Digital 1TB Raid Edition 3
    NIC Intel 10gbe CX4 Dual Port

    Switch
    2x HP 6400cl CX4

    we have 2 Citrix XEN Server as "Client"
    Hardware 2xIntel Xeon 5550W with 32GB RAM and Intel 10gbe CX4 Dual

    Performance Windows 2008

    with write cache on: ca 90000 io/s
    with write cache off: ca 25000 io/s

    Performance with Citrix XEN (windows 2008 guest):

    with write cache on: ca 66000 io/s
    with write cache off: ca 15000 io/s

    there was no differenz between raid 10 and raid 6 !

    so we use raid 6 iscsi for our production system.

    the perfomance measure was taken with sqlio and the folowing parameters

    sqlio -kW -s60 -frandom -o64 -b4 -BH -LS -Fparam.txt
    d:\testfile.dat 2 0x0 10

    for xen we make some special config.

    we also tested dss with 2x Xeon X550w and 48GB RAM , but the system seems to be slower with that config.


    so i hope this will help you.

    regards
    roger

  3. #3
    Join Date
    Apr 2009
    Posts
    62

    Default

    Thank you, I will investigate the cost of this setup,

    Which SuperMicro board did you get exactly? I see there are many models, X8DT3, X8DT3-O, X8DT3-F-O, X8DT3-F???

  4. #4
    Join Date
    Apr 2009
    Posts
    62

    Default

    Also roger, I was not able to find the Adaptec 8165BR, did you mean the 5165BR? http://www.adaptec.com/en-US/support...aid/ICP5165BR/

  5. #5

    Default

    hi drew,

    you are right. i mean the 5165br and the board is the x8dt3.

  6. #6

    Question

    rogerk, what sort of performance have you been getting with larger test file sizes? Like, twice or thrice as big as your ram?

  7. #7

    Default

    1 thread writing for 20 secs to file d:\testfile.dat
    using 4KB random IOs
    enabling multiple I/Os per thread with 64 outstanding
    buffering set to use hardware disk cache (but not file cache)
    size of file d:\testfile.dat needs to be: 41943040000 bytes
    current file size: 4194304000 bytes
    need to expand by: 37748736000 bytes
    expanding d:\testfile.dat ... done.
    using specified size: 40000 MB for file: d:\testfile.dat
    initialization done
    CUMULATIVE DATA:
    throughput metrics:
    IOs/sec: 147.68
    MBs/sec: 0.57
    latency metrics:
    Min_Latency(ms): 5
    Avg_Latency(ms): 369
    Max_Latency(ms): 7260
    histogram:
    ms: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24+
    %: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100

    this was a 40GB File from our exchange 2003 on citrix xen.
    what a performance do you aspect?

    regards
    roger

  8. #8
    Join Date
    Apr 2009
    Posts
    62

    Default

    Ouch, those results seem fairly bad to me... Especially the latency.

  9. #9

    Default

    i made another test with a non production vitual machine
    testfile 40GB

    C:\Program Files (x86)\SQLIO>sqlio.exe -kW -s20 -frandom -o64 -b4 -LS -Fparam.tx
    t
    sqlio v1.5.SG
    using system counter for latency timings, 3579545 counts per second
    parameter file used: param.txt
    file g:\testfile.dat with 1 thread (0) using mask 0x0 (0)
    1 thread writing for 20 secs to file g:\testfile.dat
    using 4KB random IOs
    enabling multiple I/Os per thread with 64 outstanding
    size of file g:\testfile.dat needs to be: 41943040000 bytes
    current file size: 10485760 bytes
    need to expand by: 41932554240 bytes
    expanding g:\testfile.dat ... done.
    using specified size: 40000 MB for file: g:\testfile.dat
    initialization done
    CUMULATIVE DATA:
    throughput metrics:
    IOs/sec: 33119.70
    MBs/sec: 129.37
    latency metrics:
    Min_Latency(ms): 0
    Avg_Latency(ms): 1
    Max_Latency(ms): 403
    histogram:
    ms: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24+
    %: 84 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1

    so i hope this help

    i think the test above was made on the wrong machine.
    our exchange has a 300GB database...

  10. #10

    Lightbulb

    Quote Originally Posted by rogerk
    i made another test with a non production vitual machine
    testfile 40GB

    C:\Program Files (x86)\SQLIO>sqlio.exe -kW -s20 -frandom -o64 -b4 -LS -Fparam.tx
    t
    using specified size: 40000 MB for file: g:\testfile.dat
    IOs/sec: 33119.70
    MBs/sec: 129.37
    latency metrics:
    Min_Latency(ms): 0
    Avg_Latency(ms): 1
    Max_Latency(ms): 403
    That's more like it!

    We need to get a 10GbE card to test with...

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •