The size of the volumes have no impact on performance at the volume level. However if there is some relationship to size and IO load, then a larger volume may have a higher IO load (IO density).
OK. Based on the 8 splits per disk recommendation, would an auto meta member size of 69905 cyl. be a good route to go? Device count is not likely to be a concern.
There is no relationship to the size of the TDATs and the size of the TDEVs.
You may need to make 8 or 16 way striped meta volumes for very high performance host LUNs, irregardless of the size of the LUN. The other option is to use several Symmextrix volumes and create a host striped volume.
The main point is that for high performance demanding volumes (1000s of IOs per second, or 100s of MB/sec) you will probably need many Symm volumes active.
Quincy561
1.3K Posts
0
December 1st, 2011 08:00
The size of the volumes have no impact on performance at the volume level. However if there is some relationship to size and IO load, then a larger volume may have a higher IO load (IO density).
BRiley1
53 Posts
0
December 2nd, 2011 06:00
OK. Based on the 8 splits per disk recommendation, would an auto meta member size of 69905 cyl. be a good route to go? Device count is not likely to be a concern.
Quincy561
1.3K Posts
0
December 2nd, 2011 07:00
There is no relationship to the size of the TDATs and the size of the TDEVs.
You may need to make 8 or 16 way striped meta volumes for very high performance host LUNs, irregardless of the size of the LUN. The other option is to use several Symmextrix volumes and create a host striped volume.
The main point is that for high performance demanding volumes (1000s of IOs per second, or 100s of MB/sec) you will probably need many Symm volumes active.