Glade with multiple "servers".

Glade with multiple "servers".

[David Brown]

Please refer me to a better place to ask questions about glade, if it
exists.

I am building a possibly distributed application (Adump, which I've
resurrected from years back <http://www.davidb.org/adump/>).  I would like
this program to be able to allow network-based backups.

A common model for this problem is to run a program on each machine that
needs to be backed up.  The machine containing the backup hardware then
connects to these machines and uses them to extract the necessary
information from that machine.  I'm trying to figure out if this model fits
well enough with the Glade or even DSA model, or if I should just implement
my own protocol over GNAT.Sockets.

The problem I see is that the notion of clients and servers is a bit mixed
around from typical.  The "clients" are generally the programs that remain
running long term, and the "server" is what is invoked periodically.  I
suppose I could have a server program that stays running, and just has the
purpose of being the boot server to maintain the connection with the
clients.

Any suggestions or advice would be helpful.

Thanks,
David Brown

Re: Glade with multiple "servers".

[Freejack]

On Fri, 16 Sep 2005 03:02:54 +0000, David Brown wrote:

> Please refer me to a better place to ask questions about glade, if it
> exists.
> 
> I am building a possibly distributed application (Adump, which I've
> resurrected from years back <http://www.davidb.org/adump/>).  I would like
> this program to be able to allow network-based backups.
> 
> A common model for this problem is to run a program on each machine that
> needs to be backed up.  The machine containing the backup hardware then
> connects to these machines and uses them to extract the necessary
> information from that machine.  I'm trying to figure out if this model fits
> well enough with the Glade or even DSA model, or if I should just implement
> my own protocol over GNAT.Sockets.
> 
> The problem I see is that the notion of clients and servers is a bit mixed
> around from typical.  The "clients" are generally the programs that remain
> running long term, and the "server" is what is invoked periodically.  I
> suppose I could have a server program that stays running, and just has the
> purpose of being the boot server to maintain the connection with the
> clients.
> 
> Any suggestions or advice would be helpful.
> 
> Thanks,
> David Brown

This model isn't entirely unusual. Look at the X Windows system, where the
server resides on the end user's workstation and the client application
resides on another workstation customized around that application.

In today's world, I've conceptually done away with the idea of
client/server communication. To me, their all just processes talking to
each other. 
Technically you could think in terms of a P2P network, where each
partition can talk to any other one, the only difference being their
assigned job. The DSA will handle this model easily. They could even
temporarily share the job if, by chance, the primary archive system goes
offline.

Now, the thing I'm a little confused about in your description of the
problem is this....Are you simply trying to back up data, or do you want
to maintain persistent+consistent processes for immediate restart should
one of them fail?

Freejack