Re: async i/o (aio and lio) desirable?

2011-10-17 Thread Alexandre Ratchov
On Sun, Oct 16, 2011 at 07:41:56PM -0400, Geoff Steckel wrote: Is there any feeling that aio_* and lio_* functionality is desirable? In theory, async i/o is very desirable in audio programs, and the lack of async i/o causes stuttering. But programs don't use async i/o APIs at all; they roll

Re: async i/o (aio and lio) desirable?

2011-10-17 Thread Piotr Sikora
Hi, The interfaces are desirable if you have code that actually uses them. Do you? I've seen almost nothing that uses these interfaces. On top of my head, both nginx and samba are using aio_* interfaces. My guess is that they're hard to use, and therefore... IMHO, they are as hard to

async i/o (aio and lio) desirable?

2011-10-16 Thread Geoff Steckel
Is there any feeling that aio_* and lio_* functionality is desirable? Do rthreads supersede them? If aio_ is desirable, what is the consensus about the tradeoff between the number of kernel thread vs. code complexity to use fewer threads to service multiple process' requests? I took a look