How to control one server with multiple clients via TCP/IP

[Opie]

I am wanting to control a single server with multiple clients.  Only one client would be active at a time, so there would be no conflict.  I want to use TCP/IP.  So far, I have programmed a cluster that passes data back to the server with no problems.  The challenge come in when a second client is added to the mix.  I have't been able to figure out how to turn each client on and send the appropriate data and then turn it off so it doesn't keep sending the same data to the server. 

Here are the things that I have considered and did some preliminary testing, but don't really know how to impliment:

1.  Send a numeric on the front of the cluster packet that tells the server that data is on the way.

2.  Send a boolean on the front of the cluster packet to somehow turn the server TCP/IP on.

The problem I have found is that LabVIEW TCP/IP doesn't like to be turned on and off.  If it doesn't get the data it expects, it goes into a reset mode and that kills the response time.

Any help?

[Opie]

Any ideas on this one?

[Jarrod_S.]

You should consider implementing a set of simple one-byte commands that can be sent back and forth between the Server and the Clients. You can base all of these ideas off the example in the Example Finder under Networking >> TCP and UDP called Multiple Connections - Server.

You will have two loops in the server VI: one to wait for new connections, and one to send and receive data from the existing connections. For instance, after one of the clients connects, it can request control of the server to send data to it by sending the character "R" for request. Every time the send/receive loop of the Server executes, the first thing it can do is to check all the existing connections to see if any of the clients have sent a control request ("R"). If so, it will create a buffer (array) of control requests. This could be in the form of Connection IDs or indexes in the array for a particular Connection ID. Your choice.

After the Server receives a request for contol, if it is not already under control by another client, then it can send a response to the first client on the control request list. For instance, the server could send the first client a "S" command for send. Note that after the clients send their control request, they should execute a TCP Read and wait indefinitely for the server to respond with the one-byte "S" command. Then, once the client in control is finished sending data to the server, it could send the character "X" telling the Server to release it from control.

The example I mentioned above already does a similar thing. Note how when a client wants to disconnect, they send the letter "Q". You can see this in the Multiple Connections - Client VI. The Server then checks each individual connection to see if it's received this one-byte command, and if it has, it closes the connection to the client. This is what you would want to implement, but instead of having just one command, you'll have to distinguish between a few and build up a buffer of control requests.

Finally, if a client does decide to disconnect in your application, they could send the command "Q" just like the example above. At this point, close the connection and remove that Connection ID from the array of connections. You will also have to handle the case that this client was in the request control waiting line when it disconnected, in which case you need to delete it from that array as well.

This will definitely work for you, but it will take some work. Best of luck!