Guidlines for distributed app developers

[Yves Male]

Hey all,

Can someone recommend a good site or book which discusses guidelines for
writing apps in distributed environment (e.g. one application across
multiple servers).

Can't use the SMP scale up concept... need to figure out rules of thumb for
scaling application out.

Topics covered would be.. when to break your exe into multiple exe's on
different nodes. How to factor Thruput, memory , app memory footprint,
etc...

Cheers,

Yves

[TLOlczyk]

On Thu, 4 Mar 2004 15:56:34 -0500, "Yves Male"
<yves_malette@raytheon.com> wrote:

>Hey all,
>
>Can someone recommend a good site or book which discusses guidelines for
>writing apps in distributed environment (e.g. one application across
>multiple servers).
>
>Can't use the SMP scale up concept... need to figure out rules of thumb for
>scaling application out.
>
>Topics covered would be.. when to break your exe into multiple exe's on
>different nodes. How to factor Thruput, memory , app memory footprint,
>etc...
>
>Cheers,
>
>Yves
>
POSA II
Doug Schmidy et al


The reply-to email address is olczyk2002@yahoo.com.
This is an address I ignore.
To reply via email, remove 2002 and change yahoo to
interaccess,

**
Thaddeus L. Olczyk, PhD

There is a difference between
*thinking* you know something,
and *knowing* you know something.

[A. G. McDowell]

In article <uJM1c.738$K6.212@bos-service2.ext.raytheon.com>, Yves Male
<yves_malette@raytheon.com> writes
>Hey all,
>
>Can someone recommend a good site or book which discusses guidelines for
>writing apps in distributed environment (e.g. one application across
>multiple servers).
>
>Can't use the SMP scale up concept... need to figure out rules of thumb for
>scaling application out.
>
>Topics covered would be.. when to break your exe into multiple exe's on
>different nodes. How to factor Thruput, memory , app memory footprint,
>etc...
>
>Cheers,
>
>Yves
>
>
I have been looking at problems related to this because some idiot
specified a requirement for failover and redundancy using a network of
PCs without RAID or shared disk, so none of our experience carries over.
So far I reckon the best we can do is to survive only if at least a
majority remain (else a network partition produces TWO systems both
claiming to be the one true system - but I need to digest the Liskov
paper further) and we have to put a copy of all state on all machines
(two successive majorities can have just one machine in common, so we
require that this machine has a copy of the saved state to carry over
the history from one majority to another). Academic writeups found
include

http://research.microsoft.com/lampso.../Abstract.html (How to
build a highly available system using Consensus - based on Paxos) and
"Practical byzantine fault tolerance and proactive recovery" by Liskov
and Castro.

I'm not encouraged by the practicalities of this. All contributions will
be gratefully received.
--
A. G. McDowell