Subject: | |
From: | |
Reply To: | |
Date: | Mon, 8 Jan 2001 07:11:47 -0700 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
This reminds me of the terminate-and-stay-resident (TSR) feature of RTE,
which would allow a program (less data segments, depending....) to remain in
memory until next executed. This was specifically for the below-stated
condition of programs that ran very often and couldn't do well with the
overhead of loading from disc each time. A good example is programming to
respond to hardware interrupts.
Is there a TSR-like feature in MPE?
Are there other aspects of MPE (other than being unsuitable for real-time
programming) that would obviate the need for TSRs?
-dtd
> -----Original Message-----
> From: HOFMEISTER,JAMES (HP-USA,ex1) [mailto:[log in to unmask]]
> Sent: Sunday, January 07, 2001 12:07 AM
> To: [log in to unmask]
> Subject: nscontrol question
>
>
> Hello Folks @ 3000-l,
>
> Re: nscontrol question
>
> ----------------------------------------------------Jeff Kell writes--
> The flip side of this is, or at least used to be, that by specifying
> a MINIMUM number of servers, these processes were 'pre-launched' by
> DSDAD so that incoming connections would quickly be farmed out to the
> waiting but as-yet-bound daemons. This avoided the process creation
> overhead and delay when establishing a new connection. In Donna's
> case above, I think preloading 80 VTSERVERs is very excessive, but I
> have been an advocate of having a minimum value > 0 since the classic
> days.
>
> Perhaps this is no longer a "negligable" gain by having a pool of
> loaded, ready-to-go daemons at hand on today's recent MPE/iX systems?
>
> The effect is not unlike APACHE preloading child processes to handle
> http requests, except that obviously the connection rate/frequency and
> persistence of said connections is an order of magnitude lower.
> ----------------------------------------------------------------------
>
> My feed back (with out testing to prove this) is that the feature of
> having available 'reserved servers' for NS-Services is marginally if
> at all useful from a performance perspective. Remember reserved
> servers are a feature from the past "MPE/VE" to be specific.
> Certainly one could think of an example or corner case where reserved
> servers would be useful - if one had an application or device(s) which
> logged on, wrote 1 record to a file and logged off several hundred
> times in a hour... but other than a case like this I.M.H.O. I really
> don't think server creation is significant when you keep in
> perspective the high end performance of the 3000 today. Not your
> fathers Oldsmobile as the commercial says!
>
> Regards,
>
> James Hofmeister
> Hewlett Packard
> Worldwide Technology Network Expert Center
> P.S. My Ideals are my own, not necessarily my employers.
>
|
|
|