back to article You're praying your biz won't be preyed upon? Have you heard of our lord and savior NVMe?

The amount of data being collected and held in systems is – yes, we know – increasing, as organisations generate and store data for real-time or post-real-time analysis. One of the drivers behind this is digital transformation. When dealing with their banks, supermarkets, and airlines, people want to experience the same slick …

COMMENTS

This topic is closed for new posts.
  1. rcxb Silver badge

    Anybody who listens to the analysis of Garner and IDC deserves to waste millions on the next trendy thing, only to find they get little benefit out of it. Look through some of their previous advice and see how that all panned out. It's hilarious!

    1. Ian Michael Gumby
      Boffin

      @rcxb

      True, but this isn't just Garner.

      Its an evolution that has been predicted.

  2. CheesyTheClown

    Why oh why

    If you’re dumping SAS anything in favor of something else, then please get a distributed database with distributed index servers and drop this crap altogether.

    Hadoop, Couch, Redis, Cassandra, multiple SQL servers, etc all support scale out with distributed indexing and searching often through map reduce methodologies. The network is already there and the performance gain is often substantially higher (orders of magnitude) than using old SAN block storage technologies.

    Or, you can keep doing it the old way and spend millions on slow ass NVMe solutions

    1. Archaon
      Facepalm

      Re: Why oh why

      Sure...because all of those scale-out database and big data platforms are well known for their ability to host virtual machine infrastructure and line-of-business applications? If you'd started harping on about something like Ceph I'd at least be able to understand - though still not entirely agree with - the relevance of the comment.

      1. Anonymous Coward
        Anonymous Coward

        Re: Why oh why

        @Archaon

        Ceph is too slow.

    2. Anonymous Coward
      Anonymous Coward

      Re: Why oh why

      Ummmmm.... without the "crap" such as SAS or NVMe... where will Hadoop, Couch, etc. store their data? Or do your servers _never_ go down? I'd agree that distributed databases and index servers are great ideas and a LOT faster, but you still need block storage somewhere.

      1. rcxb Silver badge

        Re: Why oh why

        He was insulting "SAN block storage technologies", not NVMe or SAS. His wording could have been better.

    3. Anonymous Coward
      Anonymous Coward

      @cheesey Re: Why oh why

      Even with Hadoop, you still want NVMe.

      If you can't figure this out, then you don't understand Hadoop as well as you think you do.

      PostedNVm Anon because, I'm one of the so called experts who has thought thru the problem.

      Free clue... even if you just use stand alone spark which would you rather spill to... HDD, SSD or NVMe?

      If you said anything other than NVMe then you've never had to deal with a large data set of sets of data sets.

  3. Androgynous Cow Herd

    All NVMe

    (Disclaimer - Qumulo Sales Engineer)

    Qumulo has been shipping an all NVMe NAS for over a year. Current models run ~$600/TB.

    This isn't some weird futures conversation (well, NVMeF could be risky) - but a NVMe solution presenting standard File protocols (SMB,NFS) is a really versatile solution that just requires adequate networking.

    1. rcxb Silver badge

      Re: All NVMe

      Even obscenely expensive networking gear can't come close to the speeds of good NVMe drives, let alone an array of several of them. You might get a few sales with the NVMe buzzword, but your gear is going to have glacial performance compared to internal NVMe drives. Like sticking a race car in city traffic. Of course you don't care what the article says, you're just here to spam us.

    2. Anonymous Coward
      Anonymous Coward

      Re: All NVMe

      As one comentard already pointed out you're now making the network your bottleneck.

      Detached storage is always going to be less desirable than attached storage. Cluster design is often overlooked and not well understood. I've had to go toe to toe and butt heads with hardware vendors where they don't understand that their reference architectures are wrong.

      You want to go containers... you do that over bare metal.

      NVMe if done right are always going to be faster than SSD or HDDs.

      If you talk about Hadoop, you'll want to talk about MapR's multi-MFS. (They still exist... now part of HPE)

      Posted Anon because I am an expert in this space and for more obvious reasons....

  4. IareFlash

    Anyone want a bit of cache?

    Most 21st Century storage arrays use a cache based architecture so we are already talking to the storage at NAND speeds, how the architecture handles the NVM/SCM drives is where the clever stuff comes in.

    NVMeoF is the real hero in this conversation.

This topic is closed for new posts.