Hello.
Im currently analyzing my io workflow - from what i can see in vscsistats, the mostly used length of both my read and write commands, are 4096 bytes.Most of my machines are windows terminal servers.
Now here is the big question.
Should i optimize my element size for that io size, and create (not possible, but for theoretical purposes) a disk array with a element size of 4 kb? That way, all disks will be used to read a whole stipe. Or should the element size be so big, that only one disk will be used for each io?
Default stripe size is 64 kb - and my raid arrays on my storage consists of 6 disk R10, which means a element size will be 6*64= 384 KB, correct?
Please advise !
Thanks