Conference iosg::all-in-1_v30

    Cb,
    
>A customer would like to know if it is possible to have more than one FETCHER
>running at the same time. 
    
    Don't think so.
    
>Also, is the FETCHER running as a detached process available for customers. If
>so, I would appreciate a pointer.
    
    See note 108.6 Perhaps you can sell it as EIS solution.

    Regards,
    
    Jan

Running simultaneous Fetchers may well be possible - I can't immediately think 
of anywhere which will definitely stop it:

MR should allow multiple fetches from the A1 mailbox;

I'd define OA$MTI_DATA separately for each Fetcher to ensure no clash of .NBS
filenames;

The Pending file is going to take a hammering, as both (all?!) Fetchers will 
be writing the FETCHER QUEUE record;

Set up each Fetcher as a separate ALL-IN-1 user, just as for simultaneous 
Sender and Fetcher;

Not sure if it can handle contention on OA$MTI_ERR (if any errors DO ocur!) - 
maybe this can be redefined per process as well?

Hhmm, let us know if you try it.

grahame

<
<Running simultaneous Fetchers may well be possible - I can't immediately think 
<of anywhere which will definitely stop it:
<
    I have asked this question in the past and was told it could not be
    done.  But I have done what you said and it seems to work!
    
<
<MR should allow multiple fetches from the A1 mailbox;
<
    Again, no problem that I could tell.
<
<I'd define OA$MTI_DATA separately for each Fetcher to ensure no clash of .NBS
<filenames;
<
    done.
    
    
    I ran two FETCHERs with approx 320 messages in the A1 mailbox.
    
    				Cpu Time		Elapsed time
    Fetcher 1			00:04:41.97		00:32:56.37
    Fetcher 2			00:04:40.66		00:32:47.30
    
    One fetcher with approx 320 messages in the A1 mailbox.
    
    				Cpu Time		Elapsed time
    Fetcher			00:10:31:63 		00:43:08.07
    
    This was tested on a VS3100 mod 48 16mb memory.  System load was about
    the same for both tests.  I do not know if the results would be
    diffferent on a cluster.
    
    While my tests were static, no new messages while this was running, I
    am unsure if it is worth the effort.  
    
    Any comments?
    
    
    Regards,
    Darryl

    In elapsed time alone, a saving of 25% is pretty impressive ?
    In // processing you've under halfed the CPU usage - if you were
    running on a cluster I'd expect you to see your elapsed time to half as
    well I think.   
                    
    Winton  

    Although the multiple Fetcher test "seemed" to work you will eventually
    run into problems if you run it live.

    The trouble is that each fetcher will use the FETCHER QUEUE record in
    the pending file as its temporary work queue.  This means that both
    fetchers may end up processing the same message.  This will result in
    multiple deliveries, and/or errors in the MTI_ERR log, and/or will
    cause the fetcher to bomb out completely.

    Andy

    .-1
    
    Andy,
    
>    Although the multiple Fetcher test "seemed" to work you will eventually
>    run into problems if you run it live.
    
    So the official statement to customers should be:
    
    "Multiple fetchers are NOT supported"
    
    Regards,
    
    Jan

    Dear all !

    	I'm working with Chaim and referring to the same customer's problem.

    	The customer is dealing with great amount of messages now (more
    than 3000 messages per day) and starting a universal network of VAXs
    with ALL-IN-1 . Many more messages are planned. Even though the system
    can handle regular day-to-day work, in case of network failure a large
    queue of incoming messages is created. It takes many hours (much too
    many) to receive all those messages. The customer has a grate need to
    fasten the receiving process.

    	The customer question is:
    Are there any plans to change the ALL-IN-1 fetcher so multiple
    fetchers could be run simultaneously ?

    	I'm pretty sure that more customer around the world will meet that
    problem, as we are selling ALL-IN-1 to bigger organizations.

    	I waiting impatiently for an answer, hoping desperately for a
    positive one.


    		Thanks in advance,
    		      Hanoch 

    Go to A1INFO conference for privileged information...

What is the name of this large customer and large customer
networK ?

Sounds like something you should be up the product manager
about. Send you mail to Bill Colquitt @ZKO with details.

You are right many large accounts will want a higher through.
Is what we are doing high enough ?

You tell me.

				bill

    Bill Colquitt in Notes! A commemorative stamp has been issued.