KAIO: Interesting article.

Posted in 2000

smooth1 — Tue, 8 Aug 2000 23:45:57 +0100 — source: Informix-list mailing list archive (1991-1998)

Hi guys, Found this http://www.sunworld.com/sunworldonline/swol-07-1998/swol-07-insidesolaris.ht ml though peope who are into KAIO vs using AIO VPs might be interested.... Proof, once and forall(-ish!) that at least for the latest version of Solaris that ...KAIO IS FASTER.. "5-6% on small lightly loaded systems..As the size of the system (number of processors and amount of RAM) increases, and the number of async I/O requests (load) grows, the kaio approach delivers much more scalable performance, with improvements measuring up to 30 percent under some benchmarks. Your mileage will, of course, vary. " "While the creation of a pool of worker threads up front helps provide better scalability for the user-threads async I/O facility, it still involves more overhead and a longer code path than kernel async I/O. " David. PS Why have no-one replied about the missing FAQ updates... Hello, Art , John L, the Clown...are you receiving me?? Anyone out there??? HEEEEELP!

↪ replying to smooth1

Art S. Kagel — Thu, 10 Aug 2000 12:12:59 -0400 — source: Informix-list mailing list archive (1991-1998)

Welll, I do not remember any specific submissions to the FAQ just a few "Hey, David, maybe this should go into the new FAQ!" notations on postings or responses and you said you would be searching for that kind of stuff. There was my posting about how to configure a new instance in May or so... and Obnoxio's excellent rehash of the RAID5 problem around the same time. Art S. Kagel smooth1 wrote: > > Hi guys, > > Found this > > http://www.sunworld.com/sunworldonline/swol-07-1998/swol-07-insidesolaris.ht > ml > > though peope who are into KAIO vs using AIO VPs might be > interested.... > > Proof, once and forall(-ish!) that at least for the latest version of > Solaris that ...KAIO IS FASTER.. > > "5-6% on small lightly loaded systems..As the size of the system (number of > processors and amount of RAM) increases, and the number of async I/O > requests (load) grows, the kaio approach delivers much more scalable > performance, with improvements measuring up to 30 percent under some > benchmarks. Your mileage will, of course, vary. " > > "While the creation of a pool of worker threads up front helps provide > better scalability for the user-threads async I/O facility, it still > involves more overhead and a longer code path than kernel async I/O. " > > David. > > PS Why have no-one replied about the missing FAQ updates... > Hello, Art , John L, the Clown...are you receiving me?? > Anyone out there??? HEEEEELP!