|
|
Re: Jobs API: Overhead scheduling > 100 Jobs? [message #448219 is a reply to message #448134] |
Wed, 19 April 2006 11:14 |
Benjamin Pasero Messages: 337 Registered: July 2009 |
Senior Member |
|
|
> Scheduling a hundred jobs shouldn't be a problem if there are scheduling
> rules to prevent them all running at once. The automated tests for the
> jobs framework create hundreds of jobs without difficulty. However,
> having a hundred jobs *running* concurrently would likely be a problem,
> as there is no limit on the number of worker threads that can be created
> (the VM typically runs out of memory after allocating a few hundred
> threads). Another alternative for dealing with fine-grained bits of
> work is to design a job that processes its own queue of work items. This
> approach is used quite frequently in the platform itself - for example
> there is a single job that handles a queue of viewer decoration requests.
Thanks for the reply John!
Yes, I am setting a SchedulingRule to each Job that makes sure that at
any time only 16 Jobs are running concurrently. For that I am counting
the number of running Jobs in an int field, based on a JobChangeListener.
(Is there a better way? I know there is JobManager.find(), but I rather
dont want to use that in the scheduling rule isConflicting, since its
called so often).
Its working quite well, however I am having difficulties aggregating
the progress of the running Jobs into a single Progress Monitor. My
current workaround is letting the Jobs run as system-job and having
1 Job as non-system Job running over the entire time of all Jobs and
showing Progress. This special Job is also terminating all running and
scheduled Jobs, if canceld by the user.
I know there is setProgressGroup for a Job, but from what I read from
the API, its not the right thing to use in this case, right?
Ben
> --
>
> Benjamin Pasero wrote:
>> Hi,
>>
>> I want to schedule a large number of Jobs and limit the number of
>> concurrently running Jobs to 16 using a Scheduling Rule. However, I
>> wonder if this is the right way to go for it.
>>
>> I would create all Jobs at once, schedule them and let the Scheduling
>> Rule take care of the execution.
>>
>> My alternative would be to only create new Jobs, when there are free
>> slots (less than 16 Jobs of this kind running). That would add a bit
>> more work, since I need a background-job that takes care of the
>> scheduling.
>>
>> In short: Is the Jobs API scaling well, even when 100 or more Jobs are
>> scheduled at once?
>>
>> Ben
|
|
|
|
Re: Jobs API: Overhead scheduling > 100 Jobs? [message #448242 is a reply to message #448240] |
Wed, 19 April 2006 21:46 |
Benjamin Pasero Messages: 337 Registered: July 2009 |
Senior Member |
|
|
John Arthorne wrote:
> Benjamin Pasero wrote:
>> Its working quite well, however I am having difficulties aggregating
>> the progress of the running Jobs into a single Progress Monitor. My
>> current workaround is letting the Jobs run as system-job and having
>> 1 Job as non-system Job running over the entire time of all Jobs and
>> showing Progress. This special Job is also terminating all running and
>> scheduled Jobs, if canceld by the user.
>>
>> I know there is setProgressGroup for a Job, but from what I read from
>> the API, its not the right thing to use in this case, right?
>
> I think progress groups are what you want. All jobs that share a
> progress group will show as a single entry in the progress view.
> --
I did some experiments with setting a Progress Group to these Jobs,
but was not very happy in the way progress is supported. Like, calling
"beginTask" from the 16 Concurrently running Jobs result in a very big
entry in the Progress-View (16 lines of Tasks given out). But even worse,
there was no progress indicated. Maybe this is a bug in the progress view,
since with my current workaround, I see the progress when its reported from
a single Job without a Progress Group.
Ben
|
|
|
|
Powered by
FUDForum. Page generated in 0.03025 seconds