× The internal search function is temporarily non-functional. The current search engine is no longer viable and we are researching alternatives.
As a stop gap measure, we are using Google's custom search engine service.
If you know of an easy to use, open source, search engine ... please contact support@midrange.com.



Nathan:

You observations of how the system may behave in a give situation is true, however, that does not change the base fact that there is no way to manage CPU allocation below the partition level given the current level of software. I will argue that if one of your example threads finds its way into an infinite loop (which I have seen Tomcat do on multiple occasions if the application has errors in it) then that job no matter what its priority/timeslice/shared pool is will consume as much CPU as is available until the server and/or job is forced down.

If you have Dynamic Priority Adjustment and Dynamic Priority Scheduler turned on the low priority jobs will filter up in the queue and eventually get a timeslice. If that job behaves badly and starts to consume CPU at a high rate it will continue until the job looses its timeslice or it's killed. Since most batch classes are set with a time slice of 5 seconds (5000 miliseconds) then for at least 5 seconds that thread can consume as much CPU as is available.

Since most of the systems out there are now way overpowered (meaning there is plenty of CPU to burn for the workload requirements) you will often see very high CPU percentages and that will not have affected the system in a negative way.

Jim Oberholtzer
Chief Technical Architect
Agile Technology Architects


On 4/13/2012 3:22 PM, Nathan Andelin wrote:
Not to be pedantic, but my understanding and observation is that in a mixed workload environment, when push comes to shove because a CPU is busy supporting many active threads, the high-priority threads get attention while the low-priority threads get pushed down in the stack, and wait. In James' case, It could be that Tomcat is running at a lower priority and suffering from CPU starvation while high-priority jobs are not. Tomcat's CPU requirements are also relatively high compared to traditional workloads. If it is suffering from CPU starvation, it becomes more noticeable.

Granted, threads eventually get requested CPU, but they may have wait if push comes to shove.



----- Original Message -----
From: Jim Oberholtzer<midrangel@xxxxxxxxxxxxxxxxx>
To: Midrange Systems Technical Discussion<midrange-l@xxxxxxxxxxxx>
Cc:
Sent: Friday, April 13, 2012 1:21 PM
Subject: Re: Questions about subsystems and performance (both in general and specific to Tomcat)

Time slice and job priority do not affect the amount of CPU given to a
thread, only in what order the system pays attention to the thread, and
how long the thread can be in the active queue before another thread of
equal or higher priority kicks it out. (with some of the system values
you can allow lower priority jobs to sneak in and get some attention)

Once the thread is active, it will get as much CPU as requested and
available.

Jim Oberholtzer
Chief Technical Architect
Agile Technology Architects


On 4/13/2012 12:41 PM, Nathan Andelin wrote:
> Good response, Jim. Perhaps one clarification is that you can assign job priority and time slice to sub-systems, which affect CPU allocations to servers like Tomcat.
>
>
>
> ----- Original Message -----
> From: Jim Oberholtzer<midrangel@xxxxxxxxxxxxxxxxx>
> To: Midrange Systems Technical Discussion<midrange-l@xxxxxxxxxxxx>
> Cc:
> Sent: Friday, April 13, 2012 10:46 AM
> Subject: Re: Questions about subsystems and performance (both in general and specific to Tomcat)
>
> In order:
>
> Processor is not allocated at the subsystem level, rather at the
> partition level. While there are methods to control how much memory is
> used by a subsystem and its jobs, there is no way I am aware of (at this
> point) to control processor at the subsystem level.
>
> Any subsystem will do, but you have to tune it properly. I like to set
> up routing to push specific workloads into shared memory pools where I
> can not control some of the performance.
>
> None of this is hard once you understand the job routing on the system.
>
> Jim Oberholtzer
> Chief Technical Architect
> Agile Technology Architects
>
>
> On 4/13/2012 11:14 AM, James Lampert wrote:
>> > Questions from my boss:
>> >
>> > How does one tell which subsystem has more allocated processing power
>> > and memory?
>> >
>> > And when running Tomcat, is there a particular subsystem that it should
>> > be running in, for best performance?
>> >
>> > --
>> > JHHL
>> > --
> --

As an Amazon Associate we earn from qualifying purchases.

This thread ...

Replies:

Follow On AppleNews
Return to Archive home page | Return to MIDRANGE.COM home page

This mailing list archive is Copyright 1997-2024 by midrange.com and David Gibbs as a compilation work. Use of the archive is restricted to research of a business or technical nature. Any other uses are prohibited. Full details are available on our policy page. If you have questions about this, please contact [javascript protected email address].

Operating expenses for this site are earned using the Amazon Associate program and Google Adsense.