Re: Asynchronous Execution of procedures

On Aug 2, 10:37 pm, "tom.rmadilo" <tom.rmad...@xxxxxxxxx> wrote:
On Aug 2, 4:36 am, dkf <donal.k.fell...@xxxxxxxxx> wrote:

On 1 Aug, 17:00, "tom.rmadilo" <tom.rmad...@xxxxxxxxx> wrote:

Remote execution is never going to be faster than local execution of
the same code. But if you have thousands of things to handle, it might
be faster overall to not do them sequentially.  on one machine.

Depends on the details. At work we deal with workflows that involve,
say, 100k jobs processing the same data (with varying parameters).
Each instance is reasonably computationally intensive, taking an hour
or so to execute. It's a heck of a lot faster for us to farm this out
to a cycle scavenging system

Absolutely! I was basing my comments on the observation made that
remote execution wasn't any faster than local execution, and the
number of jobs was very low, although I was unsure if the term
synchronous was confused with distributed, I don't think ecnedad
really indended to imply any synchronization between jobs, probably
just scheduling.

But if you are looking for 110k hours of work every day, it is time to
look at a distributed job scheduling system of some type.

Thank you very much for your response.
By the way I'm just giving example on the 10 number of items.
The modification will be no limit of item number.
Thank you for your explanations.

You are right, I timed the modification of one item and it took too
little time to be completed.
I just want to clarify the following:

Asynchronous Processing:

No Items: 1000
Local CPU: 4
Remote: 3

Asynchronous Processing means the process return immediately without
blocking on the calling thread.
So the 1000 Items will be processed on Local CPU 4 + 3 Remote Machines
sequentially and not definitely processing at the same time.

Is this right?

Since the processing time of one item consumed less than 1 second,
before the next item will be process to the other CPU or remote
machine, the prior item will already finished.

In this way, distributing the process of small number of items is not
wise than processing it to the Local Machine itself.

Is this right?

But in case of large number of items to be process, is it wiser to use
the distribution of processes into available Local CPU and Remote
Machines?

I'm sorry if my questions are basic, I just want to confirmed my
thinking is right.

Thank you very much and God bless (^_^)

.

Relevant Pages

  • Re: Everything is a distributed object
    ... Is it possible to send an object as a parameter to a method of a remote ... of distributing floating-point numbers across remote heterogenous ... Probably middleware services should be integrated into the language ... You can put all three layers on different machines, ...
    (comp.object)
  • Re: Queue redirection & traffic compression
    ... You need to support delegation so the client computers can access the hop to ... Downlevel Print Jobs Not Redirected Correctly ... > central and multiple remote sites. ... The print cluster queue ...
    (microsoft.public.win2000.printing)
  • Re: Scheduled Tasks Doesnt Wait for Backup to Finish
    ... Oh, by the way, all the remote jobs use UNC names. ... It works great on all the jobs that backup the local ... I could use NTBackup on the remote computer. ...
    (microsoft.public.windowsxp.general)
  • Re: Asynchronous Execution of procedures
    ... distributing the process of small number of items is not ... the distribution of processes into available Local CPU and Remote ... In your case, you have 7 processors available, so the maximum speed-up would be: ...
    (comp.lang.tcl)
  • PHP MSSQL connection problem.
    ... I have developed a PHP/MySQL web app that tracks jobs for me, ... What I what to do know is connect to the remote MSSQL server and run ... I can connect to the remote MS SQL sever with the client tools, ... What I can't do is get php to connect to it. ...
    (comp.lang.php)