I remember reading somewhere that there is a theoretical limit as to the performance for a set of disks in a disk group, meaning at some "magical" disk number the I/O (or IOPS) will begin to queue as the disk group is reading/writing as fast as possible. I don't remember where i read that, but i thought the number was relativily high (say 128 disks or so per disk group), i'm not sure if there is a consideration as to the RSS indexing or number of quorum disks within a disk group).
1) if i'm going to install a total of 240 disks in a 2C18D 8100 configuration at what point should i split into multiple FC disks groups ? (theoretical as i know i'm not going to combine FC and FATA drives).
i'm sure i'll have to dig further into the spec's of the 10k and 15k drives for the respective EVA models, but i was wondering if there is an easy way to determine the point of diminishing returns for various drive configurations on an EVA.
simply put, i've got an 2C12D EVA8100 configured with three disk groups (64x300GB10k-FC,80x300GB10k-FC,24x500GB-FATA), i'm expanding to 2C18D and i'll be purchasing 300GB FC drives to fill the additional bays, but i'm currently only approved to pickup 20x300GB disks (understanding the recommendation is to add disks in sets of 8, i'll probably pickup 4 first), the issue is that the 20x300GB drives were already received and it was noticed then that the new disks are 300GB15k-FC (main point 10k vs. 15k).
i understand that the 15k disks can be mixed with 10k disks, however, if mixed, the 15k disks will operate at "dumb'd down" 10k speeds, which is fine in small quantities (as mixing FC disk speeds is supported), but operating a large number of 15k drives in this manner negates the benefit of the additional speed of the 15k drives (and the increased $$ cost of the 15k drives).
2) is there a way to determine the max number of I/O (or IOPS) for a set of disks, trying to compare/match disk performance for 10k vs. 15k: so, for a variable of performance, say 80x300GB10k-FC, how many 300GB15k-FC disks will it take to match the 80x300GB10k-FC performance ?
the reason i'm trying to determine how many 15k spindles i'll need to match/surpass the existing 10k performance is i'm guessing the 300GB10k's will go EOL in 2009 and whether or not i should start building up a 15k disk group that won't be hampered by a small number of spindles. added to this is the understanding that for each disk group there is a loss of GB to accomodate the disk failure protection level and the 90% occupancy recommendation, the ~760GB loss on the first 8x300GB-FC is unavoidable, but at 2x300GB-FC to cover that overheard, if less 15k drives are necessary, then it's acceptable.
Another thing to keep in mind in your decision process is the applications that will be using this storage. Exchange and/or some Database applications recommend different disk groups for different "classes" of data (e.g. data vs logs); and both may be on FC disks.
