Re: To develop a cluster solution using Delphi

From: Alan Garny (someone_at_somewhere.com)
Date: 03/30/04 

Date: Tue, 30 Mar 2004 12:22:37 +0100

"Henrick Hellström [StreamSec]" <henrick@REMOVEstreamsec.se> wrote in
message news:40694f06$1@newsgroups.borland.com...
> Alan Garny wrote:
> > "Gary Williams" <2FC5DB7F@garywilliams.us> wrote in message
> > news:406834a3$1@newsgroups.borland.com...
> >>Kylix would seem to be a potential solution for that particular problem.
> > Well, considering all the bad publicity surrounding Kylix lately (i.e.
will
> > Borland keep supporting it or not?), I have no intention of using it
> > whatsoever.
> I think that is a mistake. If you have a Delphi project you have to get
> running on Linux NOW, then Kylix is an excellent solution, regardless if
> Borland ever releases any Kylix upgrade anytime in the future. Your
> worst case scenario is just that in a couple of years you either have to
> convert your project to C/C++ and keep running it on Linux, or move your
> existing code to Delphi W32/W64/.NET (whatever) and run it on windows
> instead.

Well, at the moment, my project is written in D7, so it runs on Windows.
What I am looking at, these days, is a way to get my code to run in a
parallel environment. I was originally thinking of developing a small
module, in C/C++ using MPI, that would take advantage of a cluster of Linux
boxes. Now, my view has kind of changed a bit. I, for one, would rather
stick to Windows and maybe even avoid the cluster solution and just go
straight with using the Windows computers that are on my network. The
bottleneck, as I see it, will be network communication (100 Mbits/s, instead
of 1 GBits/s as would typically be the case in a cluster). Also, I am
thinking of "simple" TCP/IP client/server solution that will be
taylored-made for my need (scientific computation). I guess I have to test
that approach to see whether it really makes sense. The two big pluses I can
see about using such an approach are:

i) Save lots time. I wouldn't have to convert my code to another language. I
would have to get MPI to work with Delphi. All I will have to do is to
create my own client/server protocol, which hopefully shouldn't take too
long, though I can already foresee some issues with "nodes" going down, etc.

ii) My users (academics) could use my software in a "parallel" environment
without having to invest in an "expensive" cluster solution. Existing
computers on a network could be used. Not to mention that no maintenance (of
the cluster) would be required, etc.

    Alan.

Relevant Pages

  • Re: How to change the heartbeat rate or should I?
    ... Communication between Server Cluster nodes is critical for smooth cluster ... each cluster network must fail independently of all other ... traffic from the network adapter that is set to Internal Cluster ...
    (microsoft.public.windows.server.clustering)
  • RE: cluster completely unavailable
    ... The network peoples says that switches was ok, network was ok, dns and wins ... Only the cluster suffered from this situation. ... completed update seq 225906 type 2 context 15 ...
    (microsoft.public.windows.server.clustering)
  • Cluster x does not appear to have a dedicated heartbeat network connection. - Unterschiedliche A
    ... nachdem ich gerade in den letzten Zuegen der Migration auf Exchange 2007 CCR ... Cluster DEFRA-EX1MBX1 does not appear to have a dedicated heartbeat network ...
    (microsoft.public.de.exchange)
  • cluster completely unavailable
    ... The network peoples says that switches was ok, network was ok, dns and wins ... Only the cluster suffered from this situation. ... completed update seq 225906 type 2 context 15 ...
    (microsoft.public.windows.server.clustering)
  • Re: Windows 2008 bizarre cluster issue
    ... Then we went through all the network ... having run out of troubleshooting items for the server configuration ... First I would try to break any and all teaming, ... Cluster servers are IBM blades ...
    (microsoft.public.windows.server.clustering)