back to article Storage newbie: You need COTS to really rock that NVMe baby

NVMe drives need NVMe fabrics to give shared arrays the data access latency killing benefits of NVMe. Unlike Nimble architect Dimitris Krekoukias, storage startup E8 believes putting NVMe SSDs in today’s all-flash arrays will be futile; it claims we need NVMe fabrics to get the NVMe performance boost. And NVMe over fabrics- …

  1. Nate Amsden

    what does that mean?

    "The idea behind E8 is that not 100 per cent of what an AFA does inside the array must be done in the AFA itself, and since NVMf requires a high-bandwidth low-latency network anyway, there will be no performance hit if those things are done outside the array"

    What are those things ? It sounds like this thing has absolutely no data services, so they get high performance(which is how Violin started??). How can some of those things like replication, or snapshots be done outside the array on a shared volume ?

    It will probably be a few years until controllers can catch up to NVMe, just like it took several years for them to catch up to regular old SSD. Maybe by 2020 ?

    Until then people will have to make due with compromises on features if they need raw performance.

    Fortunately for most customers this is a non issue since as the article says regular old SAS SSDs are plenty fast already, and will be fast enough for a long time to come.

    1. Naselus

      Re: what does that mean?

      It's pretty hard to tell tbh. The documentation available from their website isn't very forthcoming, and the other articles relating to the box are also pretty opaque. All I found were the following:

      "[a competitor won't offer] data services like thin provisioning and data reduction that E8 will supply."

      "There is no communication among the boxes themselves; no clustering or federation, and no mirroring."

      Both harvested from the previous el reg article on the company.

      So it's not out to genuinely replace your Netapps or other full-fat SME-favourite storage solutions, but will offer some dedupe and thin provisioning, and will support RAID 6. You'd be cutting back on a LOT of DR and HA functionality, but you wouldn't be completely naked on the box alone, and tbh in a DC scenario it's probably better to outsource those functions from the SAN anyway. It leaves me feeling a little leery personally (especially since they won't give it a '9's' rating, without which you can't realistically compare it's resilience to another DC class storage solution), but it's not merely a no-frills NVMeF JBOD.

    2. flashdude

      Re: what does that mean?

      From what I can gather, " things are done outside the array" might be volume management, and any other kind of data management possibly. If true, it would also mean that the host tier is contributing CPU power to generate storage performance, making this a somewhat hyper-converged solution. This makes sense since it clearly states in the article that a standard COTS server architecture like the type they use in their array on the target side won't generate this kind of performance on its own, even if you add NVMeOF on the front end and NVMe SSDs on the back end, so they chose to spread the processing to the host tier most likely??

  2. Daedalus

    OMG WTF KMN!

    This would just be an advertising puff piece if it wasn't so incomprehensible.

  3. dikrek
    Boffin

    NVMe over Fabrics is of course needed to fully realize NVMe benefits

    Hi all, Dimitris from Nimble here (http://recoverymonkey.org). Clarification:

    1. Putting NVMe drives and/or something like a 3D Xpoint DIMM in existing (modern) arrays can improve speeds, up to a point.

    2. Implementing NVMe over Fabrics is necessary to unleash the total end-to-end performance all the way to the client.

    3. Beware that speed doesn't get in the way of enterprise features, especially things that mitigate risk like multi-level checksums, snaps, clones, replication, encryption etc. Many devices out in the market are focusing on speed so much that they are ignoring even basic creature comforts.

    The challenge really is that most customers move cautiously and aren't always ready to adopt things that have barely been standardized, especially in low risk tolerance environments.

    Thx

    D

  4. abb2u

    I'm a creature that needs comfort

    Yeah, I got that need for speed -- if I'm comfortable! :-)

  5. Anonymous Coward
    Anonymous Coward

    Reliability questions

    Arent there applications that use NVME SSDs locally today? These solutions seem to bunch and share these SSDs together thus allowing for better usage of the deployed capacity without compromising much on latency. I guess the end solution will be as reliable (or unreliable) as having SSDs locally. No??

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like