Help on VP cache tunning

Hello, my friends! This is on IDS 11.50.FC6W1 on a linux_64 machine. I´m doing some vp cache optimizations, and some doubt is still un-answered: My onstat -g glo (cpu classes only) shows: class vps usercpu syscpu total cpu 7 854.44 59.08 913.52 vp pid class usercpu syscpu total Thread Eff 1 12328 cpu 195.13 14.78 209.91 746.65 28% 3 12353 cpu 226.50 15.08 241.58 727.99 33% 4 12374 cpu 169.61 10.48 180.09 581.90 30% 5 12382 cpu 111.17 7.96 119.13 438.18 27% 6 12390 cpu 74.46 5.27 79.73 354.96 22% 7 12399 cpu 58.19 4.09 62.28 289.68 21% 8 12407 cpu 19.38 1.42 20.80 150.86 13% And my onstat -g vpcache (resumed) output shows vpid pid Blocks held Hit percentage Free cache 1 12328 1707 64.9 % 97.0 % 3 12353 1753 70.2 % 80.5 % 4 12374 1602 81.0 % 87.8 % 5 12382 1423 88.5 % 93.2 % 6 12390 1871 95.0 % 92.8 % 7 12399 1871 85.1 % 93.6 % 8 12407 1019 83.0 % 94.6 % Does it seems what I´m thinking, that I can increase the cache a lot, in order to "near zero" the free cache column above? That will bring the Eff column near 100%, really??? Best regards! -- Alexandre Marini Tecnologia da Informação - DBA SEFAZ-MS / SGI-UIMP / Sistemas IBM-Informix IBM Informix Dynamic Server Certified Professional V10 / V11

You have two different onstat options below which show two completely different types of information. The onstat -g glo show about the cpu utilization. In fact that the las= t column is Efficiency which is quite low. This column gives you an idea of how well Informix and the operating system scheduling are working together. The CPU VPS should have a high efficiency. This= does not look good on this system. I would look at why the CPU VPS do not run more. The onstat -g vpcache does look very good. This indicates that most of= the memory allocation are coming from the cache. This cache is not designe= d for all allocation to come from the cache, but to have a very high number c= ome from the cache thus relieving the mutexes protecting the memory lists. John F. Miller III STSM, Support Architect miller3@us.ibm.com 503-578-5645 IBM Informix Dynamic Server (IDS) ids-bounces@iiug.org wrote on 07/08/2010 07:54:15 AM: > [image removed] > > Help on VP cache tunning [20539] > > Alexandre Marini > > to: > > ids > > 07/08/2010 07:56 AM > > Sent by: > > ids-bounces@iiug.org > > Please respond to ids > > Hello, my friends! > This is on IDS 11.50.FC6W1 on a linux_64 machine. > > I=B4m doing some vp cache optimizations, and some doubt is still un-answered: > My onstat -g glo (cpu classes only) shows: > class vps usercpu syscpu total > cpu 7 854.44 59.08 913.52 > vp pid class usercpu syscpu total Thread Eff > 1 12328 cpu 195.13 14.78 209.91 746.65 28% > 3 12353 cpu 226.50 15.08 241.58 727.99 33% > 4 12374 cpu 169.61 10.48 180.09 581.90 30% > 5 12382 cpu 111.17 7.96 119.13 438.18 27% > 6 12390 cpu 74.46 5.27 79.73 354.96 22% > 7 12399 cpu 58.19 4.09 62.28 289.68 21% > 8 12407 cpu 19.38 1.42 20.80 150.86 13% > > And my onstat -g vpcache (resumed) output shows > vpid pid Blocks held Hit percentage Free cache > 1 12328 1707 64.9 % 97.0 % > 3 12353 1753 70.2 % 80.5 % > 4 12374 1602 81.0 % 87.8 % > 5 12382 1423 88.5 % 93.2 % > 6 12390 1871 95.0 % 92.8 % > 7 12399 1871 85.1 % 93.6 % > 8 12407 1019 83.0 % 94.6 % > > Does it seems what I=B4m thinking, that I can increase the cache a lo= t, in > order to "near zero" the free cache column above? That will bring the= > Eff column near 100%, really??? > > Best regards! > -- > > Alexandre Marini > > Tecnologia da Informa=E7=E3o - DBA > > SEFAZ-MS / SGI-UIMP / Sistemas IBM-Informix > > IBM Informix Dynamic Server Certified Professional V10 / V11 > > > ***********************************************************************= ******** > Forum Note: Use "Reply" to post a response in the discussion forum.= >=

OK Mr John, thanks a lot for you explaining. Best regards.