distributed application: Networked Shared Variables, Named Services (Raw TCP/IP) or Other?

[Hillman]

Happy New Year NI forums! 

I am working on a project involving mobile interacting robots. In the future it is likely the application's components may need to run on different PCs (Targets). Note: at this point in time all the components are seperate but all running on the localhost machine. Thinking towards the future I want to pick the 'best' architecture to allow all these components (VIs performing various functions) in multiple locations. For example, several VIs on the Robots, VIs on serveral PCs. 

I am  currently aware of using Server/Client TCP/IP using named services. My mock up works well, but is it time efficient (my time coding) I wonder.. ?  

Whereas I am aware of networked shared variables which handle connections and all the parsing for the underlying tcp/ip communication. But will this be difficult the manage? I am unsure if I can associate shared variables with a VI similar to named services. I suppose I could pro grammatically create the variable upon initialization of the server component - and the client could just search the list of avaiaible variables to connect too. Downside this would require DSC module. 

As you can see, I am rather unsure. Any advice would be great!

Kind Regards,

James  

[Ben]

The lack of openess in shared variables could tie your hands latter.

I'd suggest the TCP/IP and roll-your-own since there is no signs of that going away soon.

Ben

[Jas.W]

Hi James,

You could use the Simple Messaging Reference Library (STM) to communicate over TCP/IP to increase your performance. This method makes use of creating a "meta data" message containing an index of the meta information that is sent whenever a connection is made. Therefore it has all the information needed to decode further messages. 

A more detailed explanation can be found in the following link,

http://zone.ni.com/devzone/cda/tut/p/id/4095

The link to download the library is at the bottom of that article.

If you have any further questions then feel free to post again.

Kind Regards,

Jas.W 

[Hillman]

Hi Jason,

Thanks for your reply. I hope your enjoying NI UK as much as I did.. fun times!

I have seen the link you posted a few times before. But today, I took a better look at it.

My issue is I need several multi-client severs, i.e. many servers which allow multiple clients to connect to them.

Now the STM does have an example of this - STM mutli-client Example - Server.vi (used with the STM mutli-client.vi)

However, when a make copies of these code (to have my second server) - it refuses to run. As in , it just stops itself.

I DID change the port number, on the lister aspect of the server code. But I Am unsure what else I would need to change to get this setup to work?

One thought I had was, the FIFOs all having the same name - this probably isn't a good idea between servers.

Any suggestions would be grateful!

*please could you provide me email support

Kind Regards,

James Hillman  

[Hillman]

Just wanted to quickly post - that I have resolved my problems. It was the FIFOs sharing the same name.