back to article Understanding what's going on in storage arrays is like doing MAGIC

As the announcements and acquisitions which fall into the realms of Software Defined Storage – or "storage", as I like to call it – continue to come, one starts to ponder how this is all going to work in the real world. I think it is extremely important to remember that you are going to need hardware to run this software on. …

COMMENTS

This topic is closed for new posts.
  1. Mad Chaz

    I think it's interesting to be in IT period. A lot of shifts going on right now. Back, middle and front end, as we move back to basically the old mainframe model, on a much much larger scale, in the buisness.

    What's disturbing is how it tries to force itself into consumer products in sometimes very stupid ways. Think a router that needs the internet to be configured in the first place, a little while back. I'm also thinking some people who figured it would be a good idea to have all the processing done for games on servers. Both stupid ideas.

    In the business, I'm a little worried it often ends up just leading to uncontrolled sprawls of unknown systems with no one competent taking care of them, all sitting in the cloud and getting charged to some company credit card.

    Still, all this shift in buzzwords still all sounds like the same kind of challenges we've faced before, only with new buzzwords and with less technically savy salesmen getting in the way of what's really going on.

  2. Anonymous Coward
    Anonymous Coward

    Real world concrete examples?

    Do you have an real world concrete examples of business using such DAS/SANetc in their day-to-day business?

  3. Nate Amsden

    software defined storage is looking to be as confusing as SDN

    the whole SDN space is a mess, even one of the early founders of the concept said as much.

    SDS is looking to be just as bad, if not worse. At least with networking for the most part it is stateless. If a switch goes down, or a router goes out, you may(or may not) have some downtime but the amount of risk involved(IMO) is significantly less than if say your storage resource becomes unavailable!

    It's important to keep things simple enough to manage and troubleshoot. I'm (thus far) not a fan of grid storage for transactional applications, I'm sure it's fine for object storage/archival etc. Too much risk for transactional. The software has to be really complicated.

    Storage is often striped over a large amount of resources (networking significantly less so), and so you're obviously only as strong as the weakest link. A few years ago in a more traditional enterprise class storage array I had an incident where roughly 9/300 disk drives were operating at a drive temperature that was too low (19C). While the official operating specs of the drives listed operating temps WELL BELOW 19C, they did not say the firmware had hard coded instructions to exponentially slow down the drive heads the colder the drive got.

    As a result these drives were returning 300-500ms response times. For volumes that are striped across every drive in the system(all of them) this naturally caused some problems. Most other drives were in the 20-30ms range and these were 10x that. It took some time to troubleshoot, at first I thought it was just bad drives so we tried replacing one of them to no avail. They got in contact with the manufacturer of the drive in question and they confirmed the behavior, and so the only short term course of action was to increase the temp. Fortunately the array was in somewhat of a private suite in the data center so we could increase the ambient temp in the room by about 3-4 degrees F which addressed the problem (along with closing off some cold air vents near the array). This allowed all drives to remain at 20C or above and the problem stopped. Fortunately the effects were not crippling, but they were noticeable. I believe the thermostat was set to 68 or 69 in the DC. Though the array sat upon a big steel plate which was ice cold, which may of contributed to a lower temp in the area around the array.

    I don't know about others but I never considered 19C to be "too cold", if anything I would expect significantly colder to be too cold.. but it was an interesting experience anyway.

  4. pklausner

    > Or do you come up with a method that allows a dynamic infrastructure

    > that identifies where data is located and spins/moves the compute to it?

    >

    Doesn't Joyent try exactly that with their Manta storage?

    http://www.theregister.co.uk/2013/06/25/joyent_object_store/

This topic is closed for new posts.