Hi
Thanks a lot.
I understood what I said.
However if 2nd set of data has to wait for the output of the first set
of data.
Then I am afraid it can not be done again.
Workflows will only help where both sets ( or n sets) or independent.
Have you come across any such implementation ?
In another case say when I have source code then I need to parallelize
the code.
Is Globus going to help in the parallelizing the code or the user has
to do it on his/her own.
How globus is going to be used in parallelizing? Any help/documents
are most welcome.
Thanks & Regards
JP Singh
Thanks
Quoting Nicola Pavlov <[EMAIL PROTECTED]>:
Hi again,
Well, unfortunately there is not much u can do
(parallelism) without the source code.
U can use workflows instead (CoG, Triana, Taverna,
Kepler).
Here is an example so u can understand how workflows
work:
Say, u have a job that runs at two phases. Each phase
runs over different independent sets of data. The
second phase doesn't have to wait all the time long
till the first phase finishes running on all the data
sets. It can run on the first data set after it
finishes from the first phase. In other words, without
using workflows, the first phase has to finish first
and then the second. With workflows u can achieve
pipelining.
I don't think there is another way for using globus
fatser without parallelism or workflows.
If u learn sth that i don't know, plz inform me.
--- [EMAIL PROTECTED] wrote:
Hi
Thanks a lot for your kind answer.
I just want to test the framework at this stage so I
have only 2 machines.
I have one issue what you suggested. I don't have
source code of
Nastran/TMG(IDEAS11). I had asked grid community
whether without
source code can I use globus for faster execution. I
was answered
'yes'. I am not able to understand without source
code how it can be
made parallel?
The TMG job is submitted through one INPF file. Can
it be used to
break or make it parallel.
I am wondering, what does , Globus as it is, do? Can
we use for any
practical work in its original form?
Thanks & Regards
JP Singh
Quoting Nicola Pavlov <[EMAIL PROTECTED]>:
> Hi,
>
>
> Well this question was also troubling me since i
> started to deal with grid computing. As i read
papers,
> grid promises a lot but still there is a big load
on
> users (i.e. us). In other words, we have to do
> everythin on our own.
>
> Don't want to disturb u but 2 machines can't do a
lot.
> In order to profit from grid technology u need
more
> machines. but u can try this: MPI (Message Passing
> Interface). U have to write ur jobs in (c, c++) so
> they can be paralleled. U can also try workflows.
>
> In general, u have to "model" ur jobs: inputs,
> outputs, break the jobs into smaller ones where
they
> can run independently and gain time.
>
> Hope this helped.
>
> --- [EMAIL PROTECTED] wrote:
>
>> Hi All,
>>
>> I have been able to test wsgram with a simple
job.
>>
>> I want to try some complex jobs on
>> nastran/tmg(IDEAS11) etc.
>>
>> I ran a simple tmg job however total time taken
was
>> more as compared
>> to standalone PC.
>>
>> So what I am supposed to reduce total time taken.
I
>> have grid only on
>> 2 machines.
>>
>> Thanks & Regards
>> JP Singh
>>
>>
>
>
>
>
>
____________________________________________________________________________________
> Looking for a deal? Find great prices on flights
and hotels with
> Yahoo! FareChase.
> http://farechase.yahoo.com/
>
____________________________________________________________________________________
Luggage? GPS? Comic books?
Check out fitting gifts for grads at Yahoo! Search
http://search.yahoo.com/search?fr=oni_on_mail&p=graduation+gifts&cs=bz
