What % CPU do your Prod Systems run at?

[clk430]

Just curious as to what everybody's set up is.

We have 2 CPU's @ 2800mhz RAM: 1024mb.

We changed our maxthreads from 20 to 8 and our watchmaxthreads from 0 to 8.

We are now at 100% CPU 80% of the time. I think one of our CPU's is broken. lol

 

[BocaBurger]

One map / thread can consume 100% of a cpu while it is running. If your maps are very CPU intensive, you need less maps or more CPUs. I suggest getting v. 8.0 and using it to profile and tune the paging. Even if you are not running 8.0, you can take these values and rule changes (you did use a copy of the maps and trees) and apply them to the earlier version you are using. CPU usage depends on data size and complexity, map functions and rule complexity and number of RUN and functional maps. I have seen maps that run for hours at 100% CPU and complete just fine. You just can't run 10 of them in parallel and get quick results.


BocaBurger
<===========================||////////////////|0
The pen is mightier than the sword, but the sword hurts more!

 

[clk430]

So, running at 100% 40 minutes on the hour every hour isn't a problem? It just seems that when we had maxthreads at 20 and watchmaxthreads at 0, this only happend periodically. Albeit, the server crashed quite frequently, so this prompted the change.

Nevertheless, we are heeding your advice and jumping to 8.0...but in the interim (3 months), just curious about the 100% CPU usuage in Managment concole.

 

[BocaBurger]

I would say that you have time triggered events, or a file that gets created as a trigger on a time basis. You might be able to stagger the times, or reduce maxthreads more. 2 to 6 threads per CPU would be normal, 6 threads if the CPU is very fast, however, it could be that what you are doing in the map is very CPU intensive and reducing maxthreads could help. I know of serveral companies running at 2 or 3 threads per CPU.



BocaBurger
<===========================||////////////////|0
The pen is mightier than the sword, but the sword hurts more!

 

[clk430]

You are correct. The system that get's the CPU up to 100% is an Inbound systems has 8 FTP gets, all staggered start times (in minutes) of 4,6,7,9,10,11,12 and 13. Each maps pulls from an external server, builds the files in a staging folder, then processes these X12 841. Each maps is then on a time event to trigger every 5 minutes thereafter.

You are proposing to spread the time events on these out more and/or reduce Maxthreads furthur with the above scenerio? I could give it a shot.

Thanks for all your help so far.

 

[BocaBurger]

Use the Event Server Monitor tool to see what runs when. Also check to see how long these FTPs take. By setting up the runs, to limit the number of maps running, you might also reduce network traffic and speed things up that way too. If, by running every 5 minutes you get conflicts, it might be possible to set some of the maps to run at a slower rate. Rather than stagger the start time, you can stagger the run time by setting the trigger time variables.
On a test box, see which map uses the most CPU. It should not be the FTP map, but the one that does the actual processing. The problem could also be that the FTP maps are using all the threads, and there are not enough to run the other maps, so you get into an init pending situation. The server might need more CPUs, or you may need more than one server.



BocaBurger
<===========================||////////////////|0
The pen is mightier than the sword, but the sword hurts more!

 

[clk430]

We definetly need another server or more hardware memory wise. I will do exactly what you said and test it that way. One last question - what do you mean by trigger time variables? The global triggerTime in the .ini file?

 

[BocaBurger]

NO, I was talking about the trigger time settings for the maps in the IFD.

BocaBurger
<===========================||////////////////|0
The pen is mightier than the sword, but the sword hurts more!

 

[clk430]

Got it! I appreciate your help BocaBurger.