back to article LSI gets 1 million IOPS from 1TB of flash

LSI has recorded 1 million IOPS from a terabyte of Seagate flash, showing how you need less and less flash to achieve the magic million marker. Fusion-io and IBM famously achieved a million IOPS in IBM's Project Quicksilver which came to light in September 2008. This involved 41 Fusion ioDrives with 4TB of flash hooked up to a …

COMMENTS

This topic is closed for new posts.
  1. EXAFLOPS'R'US
    FAIL

    Maths

    "Now LSI has done it with six LSISSS6200 320GB PCIe SSS cards in a single white box server running two Intel Xeon 5590 processors. That's just 1.02TB of SLC flash. "

    6 x 320GB = 1920GB in a RAID0

    Still impressive for 1.9TB of SLC goodness.

  2. Ian Michael Gumby
    Thumb Up

    But what does it cost?

    Ok, If we look at the technology, the advancements also show a clear reduction in costs too.

    One has to ask how much 6 320 GB PCIe cards cost, along with the motherboard that has 6 PCIe slots...

    This would be the more important graph to indicate the ability of companies to adopt this technology.

    Thumbs up because its pretty cool if you think about it.

  3. David Halko
    Dead Vulture

    Writer forgot to mention Sun's 1.6Million IOP F5100 Flash Acclerator

    Funny how the 1.6Million IOP Flash accelerator was not mentioned, especially when it was used to take the top spots in various benchmarks.

    http://storagemojo.com/2009/10/12/1-million-iops-in-1-ru/

    http://blogs.sun.com/BestPerf/entry/1_6_million_4k_iops

    1. Anonymous Coward
      Dead Vulture

      Also forgot...

      ... to note that Sun has introduced a few products based around flash to work with ZFS to give some significant performance gains.

  4. Karlsson
    Happy

    You left one out....

    Oracle/Sun F5100, it can do around 1.6M IOPS.......

  5. Anonymous Coward
    Anonymous Coward

    Sunacle F5100 at 1.6 million IOPS? NOT!! F5100 barely squeaks 300K IOPS

    Hard to confuse 300K with 1.6million, unless Larry's fanboys can't read their own benchmarks.

    http://www.storageperformance.org/benchmark_results_files/SPC-1C/Oracle/C00010_Sun_F5100/c00010_Oracle_Sun-F5100-Flash_SPC1C_executive-summary.pdf

    Now...300K (actual IOPS) is only 18% of 1.6million (boasted IOPS) -- but in order to get to even 18% of claimed performance, they had to 50% short-stroke (over provision) each of the 80 SSDs in this thing.

    To make matters worse -- they didn't even run a real SPC-1, but the lax'd out SPC-1C version.

    The good news is that while the F5100 may only deliver 18% of promised IOPS, it only costs $160,000 per TByte.

    Just think, at $160 per Gbyte, that's a box full of flash that's only about twice the price of an equivalent capacity of DRAM!

    Sun F5100...such a bargain.

  6. Anonymous Coward
    Anonymous Coward

    Million IOPS nonsense

    Re:

    ===========================================

    "..."None of these are benchmarks, but they do show how the amount of flash kit you need to get a million IOPS has been shrinking..."

    ===========================================

    They didn't need less flash, they merely short-stroked the Flash they used (to 50%). Funny thing is, Seagate are among those who first spotted and warned us about this deceptive practice.

    http://www.seagate.com/docs/pdf/whitepaper/standardize_storage_device_metrics_idc_wp_214418.pdf

    So...where's the SPC-1 on this thing?

This topic is closed for new posts.