I am very confused with
Low Affiniy and High Affinity, Jobs, Process etc,
My configuration is
Cognos 8.4 installed on
2 Windows 2003 Server (using as dispatchers)
Each one has 8 Gbs in RAM and 4 Quad Core Processor (equal to 16 Way i think)
Database Oracle Enterprise installed on a diferent server
We have intranet client and internet client as well.
In cognos configuration we set “Medium Configuration” for both dispatchers
But in the followings items we don’t really understand how tunning them, for example
Processing capacity (We set 16)
In the following we set (16*3= 48)
Maximum number of processes for the batch report service during non-peak period
Maximum number of processes for the batch report service during peak period
Maximum number of processes for the metadata service during non-peak period
Maximum number of processes for the metadata service during peak period
Maximum number of processes for the report service during peak period
Maximum number processes for the report service during non-peak period
In the following (we leave the default 1 to high):
Number of high affinity connections for the batch report service during non-peak period
Number of high affinity connections for the batch report service during peak period
Number of high affinity connections for the metadata service during non-peak period
Number of high affinity connections for the metadata service during peak period
Number of high affinity connections for the report service during non-peak period
Number of high affinity connections for the Report service during peak period
In the following (we leave the default 2 to low):
Number of low affinity connections for the batch report service during non-peak period
Number of low affinity connections for the batch report service during peak period
Number of low affinity connections for the metadata service during non-peak period
Number of low affinity connections for the metadata service during peak period
Number of low affinity connections for the report service during non-peak period
Number of low affinity connections for the Report service during peak period
Please any advice we will appreciate, because we have some cubes with slow performance
when they nestled dimensions.
Confusion about Low Affiniy and High Affinity requests, Jobs and Processes
Bad performing reports/cubes
The first one has to do with scaling or sizing of your Cognos Environment ie. How many concurrent requests can my server handle.
All the selected processes have to do with how many processes you allow on the server.
2 servers44 quadcores = 32 cores 2 bibus processes * 4 low affinity connections = 256 concurrent requests can be processed!
2 servers4*4 quadcores = 32 cores *2 bibus processes * 1 high affinity connections = 64 concurrent requests can be processed!
i don’t know how many users you have to support but a configuration like this 2560 active users in your environment and 25650 named users!! :o
So in theory this seems not the problem to me.
The second question has to do about designing your report and/or cube. Nesting and zero suppression is most of the time killing especially for reports in analysis studio. Most likely you have many categories in the nested level(s).
for all the properties
"Maximum number of processes for …"
Server One set the value to : 216 = 32
Server Two set the value to : 216 = 32
Am i right?
I have another question about the memory, what tuning we must to do if we have 8 gb on each Server?..
(in this moment the only thing that we did was change the IBM cognos configuration from small to medium configuration), but is that all we can do with this, what about the “Maximum number of job service connections during peak period” propertie.
Yes, but these settings will only allow you to have more concurrent processes, but your queries will still take the same amount of time to finish. for example 1.000.000 records will still 1.000.000 records after you changed the nummer of processes of the report service.
But in tuning there is an optimum and beyond that point the performance will decrease… so be carefull with allowing to many concurrent processes.
You have 8 GB of RAM for each server, so i presume you have Windows enterprise edition installed? With the Standard edition you can only use 4 GB of RAM…
The JVM memory setting (small, medium, large configuration) can be changed but you need to if it is really necessary to do so. If you set the memory for the JVM set to 1 GB, means you have 1 GB of user memory space left for your OS and all your BIBUS (report) processes. JVM blocks all the assigned memory and even if it is not used it it not usable for processes that really need it. Only on very large deployments i have seen a JVM alloaction of 700 MB…
Measure/monitor the JVM setting in real situations to see if you have to increase or decrease this setting.
The following changes can be applied in Cognos Configuration.
Change the default gateway(Cognos.cgi) to the alternate gateways (such as ISAPI, Apache_mod ). In the configuration manager the gateway URL is used only by the client tools and development tools. Change also the index.html, and default.htm in the c8\webcontent directory.
Java Memory : allocates Java Virtual Machine (JVM) for new Java objects. (max value 1024Mb)
Increase the sort memory to 16 MB. The default value is 4 MB.
Change the default font from Andale WT Unicode (International characters size of 26 Mb) to a light-weight font such as Arial (100 Kb). This change will improve PDF rendering time.
Cool topic! Tuning in growing fast… in one day lots of activity… ;D
CognosGuru - I would be careful about increasing sort memory even to 8mb as I had a bad experience back in past days with 8.2 when I was chasing random batch report failures due to out of memory (it was another one, but I don’t remember exact blurb in there, but something memory related as well). Weeks of painful WebEx sessions with Cognos support until someone in one of the chat session mention that and I lowered 8mb back to 4mb and ta-da, worked as a charm since then. Well, maybe in 8.4 it’s fixed, but I never read this explicitly in any IBM Cognos docs, so I’m trying to be careful this time… Just my 2 cents to the pile…
That is true… Tuning is 80% science en 20% art (educated guess :D). Every Cognos environment is different and you always run into other difficulties. You can tune for better and for worse… that is why a testing is an essential part of tuning. Preform some tests with the current settings change 1 setting and test again.
I have not seen this problem you describe above but it is certainly 1 i will keep in the back of my head when i encounter a ‘out of memory’ error.
you are right, the value is not fixed.
In the administration tab -> configuration tab -> dispatcher and services (menu left hand side) -> click on cognos server which is displayed as a hyperlink for example http://cognosserver:9300/p2pd
select report service properties icon
select ‘settings’ tab
look for option “Maximum number of processes for the report service during non-peak period” that is the one.
the other above this option are for the low and high affinity connections “Number of low affinity connections for the report service during non-peak period” and “Number of high affinity connections for the report service during non-peak period”
the trick is to get a benchmark first with a lot of scheduled reports. then change a setting and measure again. as long as you need to find the optimal setting for you Cognos environment.
I have a Cognos 8.4.2 Report Studio Report running in 1:13min. If I however use the same report as a source into Transformer 8.4 to build a cube it’s running in over an hour just to do the read. What could be the problem?
The datasource is Sybase IQ 15
I have read numerous post on the net and tried everything that I could. It must be some sort of setting somewhere
