Results 1 to 10 of 10

Thread: any updates RE: TRIM, RAID & LSI?

  1. #1
    I am Xtreme
    Join Date
    Jan 2006
    Location
    Australia! :)
    Posts
    6,096

    any updates RE: TRIM, RAID & LSI?

    I read somewhere that Intel will soon be able to pass TRIM commands thru to RAID arrays with their RST yaddayadda... can LSI pass TRIM commands thru yet? I'm comtemplating 2x 64GB C300/M4 on a 9240 in RAID0

    or should I not be concerned about no TRIM in RAID? (is the onboard GC in the C300/M4 good enough?)

    TIA!
    DNA = Design Not Accident
    DNA = Darwin Not Accurate

    heatware / ebay
    HARDWARE I only own Xeons, Extreme Editions & Lian Li's
    https://prism-break.org/

  2. #2
    Xtreme Member
    Join Date
    Oct 2003
    Location
    San Antonio, Texas
    Posts
    262
    been running 2 C300 64Gb in raid for about 6 months now the GC seems to be up to the task just fine. Have had no slow down in performance at all.
    Gigabyte GA-Z68X-UD4
    Intel Core I7 2600K
    2x Kingston Hyper X 240Gb's in Raid 0
    XFX 7950's in crossfire
    Custom Water Cooled
    Corsair AX850
    NZXT Switch 810



  3. #3
    Xtreme Member
    Join Date
    Nov 2011
    Posts
    124
    No TRIM or SMART passthrough for LSI in RAID yet.
    Their SAS3/SATA4 (12Gbps) controllers will hopefully bring this

  4. #4
    Xtreme Guru
    Join Date
    Aug 2009
    Location
    Wichita, Ks
    Posts
    3,887
    from speaking directly with the VP and GM of the LSI Accelerated Solutions Division (ASD) during AIS 2011 interview, it is on the roadmap, and with existing controllers. There is no official timeline. The key is the correct implementation. As testing here has shown, there are TRIM lags/hangs and other issues that can arise from TRIM. A big thing also is predictability of performance, and in massive usage flash environments, with tons of devices, the extra I/O traffic. latency concerns..etc.

    and you dont want to schedule it to happen at certain times...that can lead to unpredictable performance. OR even worse, hangs and lags like the type weve seen demonstrated here.
    In the data center predictability is the #1 concern. No deviation in performance is the goal. The introduction of TRIM brings in alot more problems than one might think jsut looking at the surface. Different write loads could lead to different levels of I/O overhead...etc...3

    most would actually prefer NO TRIM with flat line (lower) performance, than a performance gain that results in any type of deviation from SOP.
    Last edited by Computurd; 01-02-2012 at 09:45 PM.
    "Lurking" Since 1977


    Jesus Saves, God Backs-Up
    *I come to the news section to ban people, not read complaints.*-[XC]Gomeler
    Don't believe Squish, his hardware does control him!

  5. #5
    Administrator andressergio's Avatar
    Join Date
    Jul 2007
    Location
    Montevideo - Uruguay
    Posts
    5,486
    runing 3 x intel X25 G2 80GB on RAID0 since months and still same amazing performance
    Intel Core i9-7980XE@ 4.8GHz 18C/18TH (Direct Die Contact)
    ASRock X299 OC Formula
    ADATA XPG SPECTRIX D80 (4x8GB) DDR4-3800C17 B-Die
    1x Intel Optane SSD 905P 480GB
    4x HP EX950 NVMe 2TB on ASRock ULTRA M.2 CARD
    EVGA RTX 2080TI KINGPIN 2190/8000 Stock Cooling AIO 240
    SilverStone ST1500W-TI TITANIUM
    Alphacool Custom Water Cooling

  6. #6
    Xtreme Guru
    Join Date
    Aug 2009
    Location
    Wichita, Ks
    Posts
    3,887
    yes that is very true! in RAID you arent going to see that much of a difference in RAID 0, even if you are doing moderate loads. Heavy loads, YES there will be a big difference, but your talking server loads. sure you wont be doing that.

    with RAID 5/6 parity sets, etc, there will be more noticeable degradation *in my experience*

    parity sets in general also complicate the I/O overhead for TRIM. many factors there...
    "Lurking" Since 1977


    Jesus Saves, God Backs-Up
    *I come to the news section to ban people, not read complaints.*-[XC]Gomeler
    Don't believe Squish, his hardware does control him!

  7. #7
    I am Xtreme
    Join Date
    Jan 2006
    Location
    Australia! :)
    Posts
    6,096
    Quote Originally Posted by Computurd View Post
    from speaking directly with the VP and GM of the LSI Accelerated Solutions Division (ASD) during AIS 2011 interview, it is on the roadmap, and with existing controllers. There is no official timeline. The key is the correct implementation. As testing here has shown, there are TRIM lags/hangs and other issues that can arise from TRIM. A big thing also is predictability of performance, and in massive usage flash environments, with tons of devices, the extra I/O traffic. latency concerns..etc.

    and you dont want to schedule it to happen at certain times...that can lead to unpredictable performance. OR even worse, hangs and lags like the type weve seen demonstrated here.
    In the data center predictability is the #1 concern. No deviation in performance is the goal. The introduction of TRIM brings in alot more problems than one might think jsut looking at the surface. Different write loads could lead to different levels of I/O overhead...etc...3

    most would actually prefer NO TRIM with flat line (lower) performance, than a performance gain that results in any type of deviation from SOP.
    I always really enjoy your input CT!

    what would you think if, say yourself, were to contact/liaison with LSI & suggest they release TRIM support for desktop users only? - so TRIM for standalone SSDs & small RAID0/10 arrays (say ~4 SSDs max)?
    DNA = Design Not Accident
    DNA = Darwin Not Accurate

    heatware / ebay
    HARDWARE I only own Xeons, Extreme Editions & Lian Li's
    https://prism-break.org/

  8. #8
    Xtreme Member
    Join Date
    Nov 2011
    Posts
    124
    Desktop users that use LSI controllers will be a tiny drop in an ocean of Servers, I doubt TRIM will be high on the priority list.

    LSI has bought OCZ who have bought Indilinx (Octane), so LSI will have direct access to how the controller operates or is operating or is going to operate.
    No matter what happens there are going to be very interesting times ahead.

  9. #9
    Xtreme Cruncher
    Join Date
    Nov 2008
    Location
    NE Ohio, USA
    Posts
    1,608
    ^^ LSI acquired SandForce.
    24/7 Cruncher #1
    Crosshair VII Hero, Ryzen 3900X, 4.0 GHz @ 1.225v, Arctic Liquid Freezer II 420 AIO, 4x8GB GSKILL 3600MHz C15, ASUS TUF 3090 OC
    Samsung 980 1TB NVMe, Samsung 870 QVO 1TB, 2x10TB WD Red RAID1, Win 10 Pro, Enthoo Luxe TG, EVGA SuperNOVA 1200W P2

    24/7 Cruncher #2
    ASRock X470 Taichi, Ryzen 3900X, 4.0 GHz @ 1.225v, Arctic Liquid Freezer 280 AIO, 2x16GB GSKILL NEO 3600MHz C16, EVGA 3080ti FTW3 Ultra
    Samsung 970 EVO 250GB NVMe, Samsung 870 EVO 500GBWin 10 Ent, Enthoo Pro, Seasonic FOCUS Plus 850W

    24/7 Cruncher #3
    GA-P67A-UD4-B3 BIOS F8 mod, 2600k (L051B138) @ 4.5 GHz, 1.260v full load, Arctic Liquid 120, (Boots Win @ 5.6 GHz per Massman binning)
    Samsung Green 4x4GB @2133 C10, EVGA 2080ti FTW3 Hybrid, Samsung 870 EVO 500GB, 2x1TB WD Red RAID1, Win10 Ent, Rosewill Rise, EVGA SuperNOVA 1300W G2

    24/7 Cruncher #4 ... Crucial M225 64GB SSD Donated to Endurance Testing (Died at 968 TB of writes...no that is not a typo!)
    GA-EP45T-UD3LR BIOS F10 modded, Q6600 G0 VID 1.212 (L731B536), 3.6 GHz 9x400 @ 1.312v full load, Zerotherm Zen FZ120
    OCZ 2x2GB DDR3-1600MHz C7, Gigabyte 7950 @1200/1250, Crucial MX100 128GB, 2x1TB WD Red RAID1, Win10 Ent, Centurion 590, XFX PRO650W

    Music System
    SB Server->SB Touch w/Android Tablet as a remote->Denon AVR-X3300W->JBL Studio Series Floorstanding Speakers, JBL LS Center, 2x SVS SB-2000 Subs


  10. #10
    Xtreme Guru
    Join Date
    Aug 2009
    Location
    Wichita, Ks
    Posts
    3,887
    Desktop users that use LSI controllers will be a tiny drop in an ocean of Servers, I doubt TRIM will be high on the priority list.
    yup i agree. Even if i begged they wouldnt do it

    alot of it is also validation, they would never release anything to the public that hadnt been fully validated, if it got out in the wild...
    "Lurking" Since 1977


    Jesus Saves, God Backs-Up
    *I come to the news section to ban people, not read complaints.*-[XC]Gomeler
    Don't believe Squish, his hardware does control him!

Bookmarks

Bookmarks

Posting Permissions

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