Best SATA drive for FreeNAS SLOG - Intel S3710 or S4600?

Discussion in 'FreeBSD and FreeNAS' started by lunadesign, Mar 9, 2018.

  1. T_Minus

    T_Minus Moderator

    Joined:
    Feb 15, 2015
    Messages:
    6,404
    Likes Received:
    1,313
    yep! Funtin adapter should work fine.

    I'm about to try one on an Intel E5 v1/2 board and see how it improves some drives :)
     
    #21
  2. modder man

    modder man Active Member

    Joined:
    Jan 19, 2015
    Messages:
    618
    Likes Received:
    59
    Oh well, let us know how that goes. That was my plan.
     
    #22
  3. BackupProphet

    BackupProphet Well-Known Member

    Joined:
    Jul 2, 2014
    Messages:
    720
    Likes Received:
    253
    I use Optane 32GB as ZIL on a PCIe M.2 adapter on a SuperMicro X8 motherboard. Works very well. 200x better sync write performance :) I am on FreeBSD 11.1-RELEASE.
     
    #23
  4. fohdeesha

    fohdeesha Kaini Industries

    Joined:
    Nov 20, 2016
    Messages:
    874
    Likes Received:
    638
    I was planning on doing the exact same on a new ZFS build, but this STH article and others: https://www.servethehome.com/intel-optane-memory-m-2-16gb-32gb-work-servers-zfs-cache-devices/

    warn that it's very slow for an NVMe drive and more importantly has incredibly poor write endurance - 100GB/day - they do NOT recommend them for use as an SLOG device

    not sure what else to look at, I don't do many sync writes so I don't want to dump $400 on a big boy Optane, but would still like something small there to reliably boost sync writes out of the sub 5MB/s range

    EDIT: just realized the 58GB Optane 800P is only $100, and has much better write endurance, guess I'll go with that
     
    #24
    Last edited: Jul 4, 2018
  5. BackupProphet

    BackupProphet Well-Known Member

    Joined:
    Jul 2, 2014
    Messages:
    720
    Likes Received:
    253
    What I do is turn off sync write for areas there is no need, for example ISO files of different Linux/Windows distributions. But for databases or virtual machines, having sync enabled is a good idea.

    It is not slow at all, performs better than my Intel S3700. Endurance may be an issue, but it is a cheap start until the big Optane drives becomes more affordable.
     
    #25
    fohdeesha likes this.
  6. svtkobra7

    svtkobra7 Active Member

    Joined:
    Jan 2, 2017
    Messages:
    174
    Likes Received:
    34
    Optane! Interesting thread here ATM : SLOG benchmarking and finding the best SLOG

    Optane 280GB (as 20GB vDisk) v. S3700 100GB = Optane less than 1/10 of the latency and almost 10x the sync write speed at 8192 kbytes.

    Optane 900p - 280GB (20GB vDisk)
    Code:
    diskinfo -wS /dev/nvd2
    /dev/nvd2
              512             # sectorsize
              21474836480     # mediasize in bytes (20G)
              41943040        # mediasize in sectors
              0               # stripesize
              0               # stripeoffset
              Virtual disk    # Disk descr.
              VMWare NVME-0001        # Disk ident.
    
    Synchronous random writes:
               0.5 kbytes:     45.2 usec/IO =     10.8 Mbytes/s
                 1 kbytes:     47.6 usec/IO =     20.5 Mbytes/s
                 2 kbytes:     46.5 usec/IO =     42.0 Mbytes/s
                 4 kbytes:     44.6 usec/IO =     87.5 Mbytes/s
                 8 kbytes:     47.0 usec/IO =    166.2 Mbytes/s
                16 kbytes:     49.8 usec/IO =    314.0 Mbytes/s
                32 kbytes:     58.6 usec/IO =    533.5 Mbytes/s
                64 kbytes:     76.4 usec/IO =    818.3 Mbytes/s
               128 kbytes:    114.6 usec/IO =   1091.1 Mbytes/s
               256 kbytes:    178.9 usec/IO =   1397.5 Mbytes/s
               512 kbytes:    294.3 usec/IO =   1699.2 Mbytes/s
              1024 kbytes:    527.2 usec/IO =   1896.7 Mbytes/s
              2048 kbytes:    981.7 usec/IO =   2037.3 Mbytes/s
              4096 kbytes:   1896.4 usec/IO =   2109.2 Mbytes/s
              8192 kbytes:   3739.1 usec/IO =   2139.6 Mbytes/s
    S3700 - 100GB
    Code:
    root@Carmel-SANG2:~ # diskinfo -wS /dev/da7
    /dev/da7
            512             # sectorsize
            100030242816    # mediasize in bytes (93G)
            195371568       # mediasize in sectors
            0               # stripesize
            0               # stripeoffset
            12161           # Cylinders according to firmware.
            255             # Heads according to firmware.
            63              # Sectors according to firmware.
            ATA INTEL SSDSC2BA10    # Disk descr.
            BTTV3452001M100FGN      # Disk ident.
            id1,enc@n500304800139d9fd/type@0/slot@1/elmdesc@Slot_01 # Physical path
            Not_Zoned       # Zone Mode
    
    Synchronous random writes:
             0.5 kbytes:    301.8 usec/IO =      1.6 Mbytes/s
               1 kbytes:    294.7 usec/IO =      3.3 Mbytes/s
               2 kbytes:    265.3 usec/IO =      7.4 Mbytes/s
               4 kbytes:    176.9 usec/IO =     22.1 Mbytes/s
               8 kbytes:    191.4 usec/IO =     40.8 Mbytes/s
              16 kbytes:    235.7 usec/IO =     66.3 Mbytes/s
              32 kbytes:    300.8 usec/IO =    103.9 Mbytes/s
              64 kbytes:    433.2 usec/IO =    144.3 Mbytes/s
             128 kbytes:    686.3 usec/IO =    182.1 Mbytes/s
             256 kbytes:   1265.0 usec/IO =    197.6 Mbytes/s
             512 kbytes:   2489.4 usec/IO =    200.8 Mbytes/s
            1024 kbytes:   4998.6 usec/IO =    200.1 Mbytes/s
            2048 kbytes:  10029.3 usec/IO =    199.4 Mbytes/s
            4096 kbytes:  20055.8 usec/IO =    199.4 Mbytes/s
            8192 kbytes:  40179.3 usec/IO =    199.1 Mbytes/s
    
     
    #26
  7. BackupProphet

    BackupProphet Well-Known Member

    Joined:
    Jul 2, 2014
    Messages:
    720
    Likes Received:
    253
    The most interesting numbers here are the difference at
    4kb which is a good default for ZVOLs, especially SQL Server demands 4kb record size and refuse to work on larger record sizes. Because of this you also hardly get a benefit from the transparent ZFS compression :(
    8kb which is the typical recordsize for NFS and the default block size for databases like PostgreSQL and Oracle
    16kb is the default block size for MySQL

    I usually give PostgreSQL a record size of 32kb as it will compress a lot better, and I dont have a heavy OLTP workload. If you have a heavy OLTP workload you should use a 8kb recordsize.

    The other sizes are a lot less interesting.

    EDIT: Actually recent NFS clients use an optimized block size for communication
     
    #27
    Last edited: Jul 6, 2018
  8. Rand__

    Rand__ Well-Known Member

    Joined:
    Mar 6, 2014
    Messages:
    2,599
    Likes Received:
    343
    VSAN uses 64KB slices so interesting for moving between that and ZFS storage imho :)


    also that might actually depend on nfs if I think about it ... makes me wonder whether nfs dynamically adapts...
     
    #28
  9. anomaly

    anomaly Member

    Joined:
    Jan 8, 2018
    Messages:
    136
    Likes Received:
    14
    Nobody mentioned but: SAS3 SSD is another option :)
     
    #29
Similar Threads: Best SATA
Forum Title Date
FreeBSD and FreeNAS How best to determine what to upgrade for better FreeNAS performance? Oct 1, 2018
FreeBSD and FreeNAS FreeNAS: Best RAID for my array? Apr 21, 2018
FreeBSD and FreeNAS What's the best alternative to FreeNAS? Oct 14, 2017
FreeBSD and FreeNAS ZFS build questions. How to best use my current drives? Sep 22, 2015
FreeBSD and FreeNAS Is FreeBSD the best way to go? May 22, 2012

Share This Page