If you have enough memory (which you should have, in order to unleash all the available power in the processors ;-)) you can increase the number of aioservers to 100 per CPU or more. You have to increase maxreqs also because it doesn't help having free aioservers, if there's no free AIO request buffer space...
If you do a
pstat -a|grep aioserver (could be pstat -A - don't remember exactly)
and notice that they are all consuming CPU seconds, then you don't have enough of 'em. If you have some which are at 0:00 CPU seconds, then you're fine. If you have a lot at 0:00 then you can go down.
The trouble with not having enough aioservers is, when they are all busy and the application needs to do an async IO, the application will go into WAIT until an aioserver gets freed. So instead of having an aioserver doing all the waiting for IO, your application also gets delayed.
The trouble with having too many aioservers is that you are eating away at memory that may have other more efficient use.
Limiting the number of aioservers on dev or test machines is also a way of limiting the load on your SAN server(s) by those dev/test machines.
HTH,
p5wizard