back to article It pays to fake it: Test your flash SAN with a good simulation

It is pretty obvious that storage systems vary. You could reply, with some justification: “No shit, Sherlock!” What is less obvious and more useful to know, however, is how and why they vary and how the variation – not just between all-disk, hybrid and all-flash arrays but even between different arrays of the same class – can …

  1. Nick Dyer

    Nick Slater makes the most important point of the whole article

    "Flash storage has great performance benefits for random workloads over spinning disks, but it’s still relatively expensive and has no performance value when it comes to sequential write workloads".

    Well said that man.

    Whilst there's huge amount of hype on flash/SSD storage silos for accelerating performance, the biggest misconception is that any workload will be lightning(!) fast on flash... whereas any sequential workload (writes, or reads) can actually be outperformed by using higher capacity spinning rust at a fraction of the price.

    But of course, any good marketing veep will never mention that if the desire is to shift as much flash as possible!

    1. Tom Maddox Silver badge
      Meh

      Re: Nick Slater makes the most important point of the whole article

      I actually prefer unicorns for servicing my all-sequential workloads, since they also don't exist. Under real-world use conditions, I have never found a use case where flash storage in an array designed to make proper use of it was not helpful. Now, some vendors are more effective than others at making use of flash storage, but that's a different issue.

      1. Sixtysix

        Re: Nick Slater makes the most important point of the whole article

        I do agree that proper configuration of a flash enabled SAN is critical, but I suspect the argument and questions that should arise are actually different, and whilst I don't want to cut into the used Unicorn Market, I think there is real merit in taking the original direction of thought further.

        I can appreciate that most current "real world" scenarios may not fit the "all sequential" scenario, but that is no reason not to realise the issue, and plan for it.

        If your server stores multiple types of data - mixtures of little files, big files, massive files, audio, video, and who knows what else, why does it all need to live in one file system/virtual disk/LUN/container/whatever. The take-away here is that the small and frequently used stuff should be homed on electric silicon, and the big files and steaming stuff tucked into centrifugal storage. Analyse, model and break down the storage needs by use case to make sure you can best use the capabilities of the storage facilities you've got available!

        Yes, that is what a top range well configured and specified SAN will do on the fly, but similar benefits can be realised without top dollar with some forethought. Horses for courses no less.

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