KAIO questions
I'm looking for comments on the reliability of KAIO versus performance in IDS 11.7. I cannot seem to get any quantitative number on what it gives IDS over regular IO. I also can't find anything discussing risk factors of async IO.
IO performance can vary greatly, but in general I have seen KAIO to be equal or slightly faster than Informix AIO. The real benefits has been in less cpu per I/O operation. John F. Miller III STSM, Embedability Architect miller3@us.ibm.com 503-578-5645 IBM Informix Dynamic Server (IDS) ids-bounces@iiug.org wrote on 06/09/2011 12:55:06 PM: > [image removed] > > KAIO questions [23967] > > JARED HEATH > > to: > > ids > > 06/09/2011 12:55 PM > > Sent by: > > ids-bounces@iiug.org > > Please respond to ids > > I'm looking for comments on the reliability of KAIO versus performance in IDS > 11.7. > > I cannot seem to get any quantitative number on what it gives IDS > over regular > IO. I also can't find anything discussing risk factors of async IO. > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. >
Jared, No risk that I am aware of. I've been using KAIO in Informix for many years without issue. I did some benchmarks many years ago, and updated with information I've gotten from other folk about the behavior of DIRECT_IO, here's what I know performance wise: 1. RAW device disk chunks with KAIO enabled is fastest 2. COOKED filesystem file chunks with DIRECT_IO enabled, and using KAIO, is about 3-5% slower than #1 3. COOKED device disk chunks with KAIO enabled are next at about 8-10% slower than #1 4. COOKED filesystem file chunks without DIRECT_IO (and so no KAIO available), is 5-15% slower than #3 depending on the FS type and your IO patterns (this option is VERY dependent on the speed of your system's memory used for the OS buffer cache - hence the wide range) I have not done much testing of RAW and COOKED device chunks without KAIO. The one system I used that on did not support KAIO in the OS (IDS simulated it badly/slowly - those were old Data General systems) so no comparison was available. Art Art S. Kagel Advanced DataTools (www.advancedatatools.com) Blog: http://informix-myview.blogspot.com/ Disclaimer: Please keep in mind that my own opinions are my own opinions and do not reflect on my employer, Advanced DataTools, the IIUG, nor any other organization with which I am associated either explicitly, implicitly, or by inference. Neither do those opinions reflect those of other individuals affiliated with any entity with which I am affiliated nor those of the entities themselves. On Thu, Jun 9, 2011 at 3:55 PM, JARED HEATH <jared.heath@sbcglobal.net>wrote: > I'm looking for comments on the reliability of KAIO versus performance in > IDS > 11.7. > > I cannot seem to get any quantitative number on what it gives IDS over > regular > IO. I also can't find anything discussing risk factors of async IO. > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --bcaec51ba4b14609f504a55cf257