> So how is it possible that reads of d11,d12 sum up in
> d10, but writes of d20, d21, and d22 are all equal.
> Is this a bug?

No sir. This is how RAID 1+0 theory works. You wrote yourself:

"reads of d11,d12 sum up in d10,"

"but writes of d20,d21 and d22 are all equal."

The first difference that jumps out at us here is reads versus writes.
Unless one did some tuning with `metaparam`, default behavior for reads is 
round robin.
What that means is that reads will be load balanced across all submirrors; so 
if one disk delivers 25MB/s, reads on a mirror will deliver 50MB/s.

Writes are different. Why are they different? Because for every write I/O that 
goes out to the
mirror (bulk or otherwise), it has to be written out as many times as there are 
submirrors, since that is what mirroring is all about.

Since that is the case, no load balancing can take place. Thus, a write on a 
mirror with n submirrors can only be as fast as the speed of every individual 
disk in that mirror, since write I/O has to happen for every disk. That is 
under ideal conditions, writes are normally slightly slower.
 
 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to