HPC

DDN claims burst buffer bashes 'past 1TB/sec bandwidth'

Japanese superdupercomputer kept fed with data by burst buffer beast


This is a we-can-pee-up-the-wall-higher-than-anyone-else story – one which we'd normally give a miss - only the numbers are past 1TB/sec – head-scratchingly high.

The Oakforest-PACS T2K massively parallel 25 petaflop superdupercomputer machine is number 6 in the TOP500 list at SC16 in Salt Lake City. It's operated by the Joint Center for Advanced High Performance Computing (JCAHPC), which is run collaboratively by the Information Technology Center at the University of Tokyo and the Center for Computational Sciences at the University of Tsukuba.

DDN's IME14K (Infinite Memory Engine) arrived in November 2015 as a flash cache between supercomputing compute nodes and parallel file system storage. It was given 7.68TB Toshiba NVMe flash drives in August.

The Japanese installation's file system cache is a monster, comprised of 25 x DDN IME14KX systems, which take up two-and-a-half data centre racks or more. Each IME 14KX has 48 x 800GB NVMe SSDs, and eight Intel Omni-Path ports. That makes 1,200 SSDs in total.

The 25 systems provide a cache of 960 TB (25 x 48 x 800GB) and a logical bandwidth of 1.5 TB per second, meaning 1,250MB per drive. There are several faster NVMe drives out there but it may be they would saturate the IME14KX IO system.

DDN claims other methods of getting past 1TB/sec with file systems need 25 to 250 racks, 10 to a 100 times more racks.

DDN tells us that it measured the machine's I/O performance by using the Livermore Computing Center's IOR benchmark, which looks at two IO access patterns:

  1. FPP - File Per Process where parallel processes perform independent file IO respectively
  2. SSF - Single Shared File where all parallel processes perform I/O to a single shared file

The benchmark results were:

  • FPP write and read at 1.14TB/sec and 1.2TB/sec respectively
  • SSF write and read at 1.18TB/sec and 1.25TB/sec respectively

That's not just peeing up the wall; it's doing a reverse Niagara with aggregated fire hoses.

DDN claims "SSF is an access method that cannot realise sufficient performance with a conventional parallel file system, but is considered an effective access method for the next generation of Exascale supercomputers."

So there's DDN set fair for exascale supercomputing storage. ®

Similar topics

Broader topics


Other stories you might like

  • Robotics and 5G to spur growth of SoC industry – report
    Big OEMs hogging production and COVID causing supply issues

    The system-on-chip (SoC) side of the semiconductor industry is poised for growth between now and 2026, when it's predicted to be worth $6.85 billion, according to an analyst's report. 

    Chances are good that there's an SoC-powered device within arm's reach of you: the tiny integrated circuits contain everything needed for a basic computer, leading to their proliferation in mobile, IoT and smart devices. 

    The report predicting the growth comes from advisory biz Technavio, which looked at a long list of companies in the SoC market. Vendors it analyzed include Apple, Broadcom, Intel, Nvidia, TSMC, Toshiba, and more. The company predicts that much of the growth between now and 2026 will stem primarily from robotics and 5G. 

    Continue reading
  • Deepfake attacks can easily trick live facial recognition systems online
    Plus: Next PyTorch release will support Apple GPUs so devs can train neural networks on their own laptops

    In brief Miscreants can easily steal someone else's identity by tricking live facial recognition software using deepfakes, according to a new report.

    Sensity AI, a startup focused on tackling identity fraud, carried out a series of pretend attacks. Engineers scanned the image of someone from an ID card, and mapped their likeness onto another person's face. Sensity then tested whether they could breach live facial recognition systems by tricking them into believing the pretend attacker is a real user.

    So-called "liveness tests" try to authenticate identities in real-time, relying on images or video streams from cameras like face recognition used to unlock mobile phones, for example. Nine out of ten vendors failed Sensity's live deepfake attacks.

    Continue reading
  • Lonestar plans to put datacenters in the Moon's lava tubes
    How? Founder tells The Register 'Robots… lots of robots'

    Imagine a future where racks of computer servers hum quietly in darkness below the surface of the Moon.

    Here is where some of the most important data is stored, to be left untouched for as long as can be. The idea sounds like something from science-fiction, but one startup that recently emerged from stealth is trying to turn it into a reality. Lonestar Data Holdings has a unique mission unlike any other cloud provider: to build datacenters on the Moon backing up the world's data.

    "It's inconceivable to me that we are keeping our most precious assets, our knowledge and our data, on Earth, where we're setting off bombs and burning things," Christopher Stott, founder and CEO of Lonestar, told The Register. "We need to put our assets in place off our planet, where we can keep it safe."

    Continue reading

Biting the hand that feeds IT © 1998–2022