Re: Kernel Asynchronous IO on HP-UX

Defect 92132 is one such reported problem: It reads thus: HP KAIO HAS A SIGNIFICANT I/O BOTTLENECK WHICH CAN CAUSE 30% TO 75% IN PERFORMANCE The general symptom of the problem is that HP KAIO (native kernel async I/O) is *much* slower than Informix AIO. The impact is anywhere from 30% to >100% slowdown in disk I/O. The problem lies in the server's I/O polling mechanism. When the server issues an I/O request through KAIO, it must poll the HP-UX asynchronous driver to determine when the request has been fulfilled. If a request has not yet been filled, the I/O thread may choose to go to sleep for 10 milliseconds before checking the request status again. Up through ODS 7.1, the I/O thread would sleep using a select() system call. This was changed in 7.2 to use the newer and lighter nanosleep() system call. Performance tests showed this was lighter on the CPU, and still gave us the desired sleep behavior. Sometime after 7.2 was released (we believe), nanosleep() was changed to correctly conform to its IEEE standard. One of the impacts is that we can no longer sleep for just 10 milliseconds. On average, the I/O thread is now sleeping for almost twice that long (when it does not need to) and this is why we believe KAIO performance has gotten so much worse. I know of one of customer's running into this when upgrading from 7.13 to 7.24.UC5. We don't use KAIO on our HP kit because of the performance hit. I'm not aware of any release which has fixed this, either HTH Pete Smith DBA Aethos Communication Systems EsquivelR wrote: > I have two Unix servers which run Online 7.23.UC7 and Online 7.14.UD6. Both > servers are running HP-UX 10.20. Has anyone out there had any experience using > Kernel Asynchronous IO? I'm looking for any known "gotchas" or testimonials on > increased performance. > > esquivelr@aol.com