Re: Bloody ERP schema

ART IS GOD!!! Well not really - but he knows of what he speaks. Take it as Biblical advice and run with it. He's never steered me wrong. Ok there is an occasional syntax error but hell the scripts are free and who can test every script on every platform. ART RULES!!! Rob Vorbroker > On Tue, 20 May 2003 07:22:22 -0400, Oxow wrote: > >> Hi, it is me again. >> >> For you information, a Aix specialist has studied the behaviour of the >> box and he told us that the machine was waiting after informix with >> (vmstat, top, etc..). He also noticed that several informix jobs were >> sleeping with the command onstat -g ses <session number>. >> >> Concerning the NETTYPE, this one was recommended by an IBM/Informix >> guy working with the ERP company, he explained us - when he advice us >> to modified these parmeters - that as we had several servers defined >> for x clients connected to the database server, it was preferable to >> comunicate with network listeners instead of shm one. > > The IBM/Informix guy is wrong, period. ESPECIALLY if most of your > connections are through TCP/IP then tcp listeners MUST be run in NET VPs > not in CPU VPs! As someone mentioned, I posted a rather long winded > explanation of this about a month ago which has prompted a reexamination > of the manual's recommendations within IBM. They are testing my > assumptions and recommendations now. Meanwhile, dozens of sites are > running better for having take the advice. TCP listeners in a few NET > VPs and shm poll threads in EVERY CPU VP (so 8 in your case) for best > user response time and minimized system impact. > >> Fot the NOAGE parameter, as I wrote it above the Aix machine is not >> overloaded by the work so I do not think so that informix take the >> entire resources of the box. In addition, even if informix would take >> a majority of the ressources, it will be what we would desire. Like >> this, it will be easier to fix the situation. So that is the reason >> why the NOAGE parameter is set to 0. > > It would seem to me that while AIX is not the most aggressive OS at > aging long-running processes (that would be HPUX) keeping the engine's > priority high by setting NOAGE 1 could only help engine performance and > if as you say the engine is not busy enough to affect overall system > performance, what's the harm? > >> Now, for the AFF_SPROC and the AFF_NPROCS I will thinking on it. >> Thanks >> >> Finally, for what it is the update statistics, even if I run it with >> the high level I do not understand why if I run it when nobody is >> working on the database it coould be the core of my problem ? Please >> explain me that. >> In addition, I thought that the update stats did not take a lot of >> ressources? > > I think it's not the resources of running the stats that folk are on you > about. The recommended levels, as implemented by dostats and documented > in the Performance Guide, are normally sufficient for the optimizer and > are cheaper to produce that full HIGH runs against every column and much > cheaper than a HIGH at the table or column level. But if you are doing > the HIGH only at the table or database level know that while it provides > data distributions that are AT LEAST as good as those produced by > dostats, you may not be setting the low level index statistics properly > which may cause the optimizer to select a sub-optimal query plan. > Either switch to performing the recommended suite of commands; either > with dostats, one of the other utilities that implement this protocol, > or manually; or additionally run a LOW on the entire key of each index. > >> Hoping that I am clear enought. > > You are being clear, but you are also being negative. You asked for > help then methodically listed reasons why you do not want to implement > any of the suggestions. Give it a try for cripes sake. You must have > thought some of us know what we're doing or you would not have posted! > > Art S. Kagel

Hi, I will as soon as I could test the recommandation of Mr. Art Kagel and the other separatly on a smaller Aix box. And I will post the results and my impressions of what it will take place in order to help people who could have the same problem than us. Anyway, if you have other suggestions. Do not hesitate to share it. For All of you have done for us, I would like to sincerly thank you. Best Regards Oxow > ART IS GOD!!! Well not really - but he knows of what he speaks. > Take it as Biblical advice and run with it. He's never steered me wrong. > Ok there is an occasional syntax error but hell the scripts are free and > who can test every script on every platform.