I got notification today about the new disks available for the USP et al.
I thought it interesting that there were a couple of things that might need comment.
Even though it was not really suggested, the notificaiton said:
…. due to the large capacity and the proportionally longer rebuild time for these drives, a RAID-6 configuration provides additional protection against possible data loss in the event of a second drive failure during the longer rebuild time of the first failure.
I like the idea that someone actually says it is a good idea as so far, no one is really pushing that. Also considering our double disk failure in one parity group, it is something to think about.
Now comes the hard bit. As the drives get bigger, the IOPS stay much the same. I would expect that the 15k 300 GB drives would have similiar IOPS to the 15k 146 GB drives and you would have to consider that if you wanted to go that path. It also leads me to something I read that states:
Each Array Control Pair (ACP) on the Universal Storage Platform can support 15,500 IOPS. Thats about one third of the IOPS the actual disks can produce.
I was lucky to have done a lot of maths with my Computer Science degree and if we are to use all those recommendations about such and such a configuration, I might actually be able to use what I learnt. I will have to dust off my calculator and probably replace its batteries. I can hardly wait until we can only get minimum 500 GB drives that spit out 100 IOPS max. Then the ACP's will have enough back end disks to just cope.
Actually I am happy to use the 300 GB 15k drives if they are priced well and HDS drop the licencing costs per TB. I wanted HDS to provide 500 GB SATA disks for the USP so I could provide cheaper backup devices to our servers. EMC can now do it.. Its all about getting the right solution for the right price.