如何控制在多个端口上侦听的Java服务器应用程序

Question

First, forgive me if this has been answered already. I spent a few hours searching the Internet (including stackoverflow) for my answer/good tips but what I am finding is not exactly what I am looking for so I decided to post...here goes.

I am writing a program to receive a "feed" over sockets on a single (Windows) machine. I will have 6 feeds coming in so I am listening on 6 ports (7000-7005). Each "feed" will be certain data I want to read/parse and write to a DB. The data over each socket will be sent to me every 15 seconds (eg data is refreshed every 15 seconds).

I have written a simple java program (modeled as a server program) that will simply listen on the specified port, receive the updates and write them to the DB. Let's generically call this the "listener" program.

I can change the port at the command line so what I've done in this early stage is to open 6 command prompts and I run java -jar myprogram.jar and I see what I expect to see. My understanding of this setup is that I have 6 JVMs running (basically 6 separate processes).

My question revolves around controlling each listener. I want to create an admin interface for the listeners...I want to be able to control the starting/stopping of each listener/port combo from a CENTRAL script/program.

I've worked with Threads before and thought to modify my existing app to start and create 6 threads to handle the listening for each port. From the main program I can then invoke commands to control the child threads by port number or some name.

My thought/concern for that design is that what if I get an update for any feed where some significant change has occurred AFTER it's thread had "execution focus"? Threads are time-sliced usually right? How to I ensure that these child threads are always ready to get the updates?? I know CPUs switch fast :-) but isn't there SOME (albeit small) possibility that I miss an update by virtue of the time-slicing I'd think?

So then I thought to keep it as-is and just run multiple instances of the program where each one is bound to a specific port. Ok..well if I do that, how am I controlling all instances from a single place? I've read about RMI a little..is that the answer? So I could call "remote" methods from a control program against the other instances of the server application.

I am looking for some tips/best-practice for this problem scenario. Hoping I've explained what the issue is clearly enough. Thanks for taking a read and for posting any thoughts/links, etc.

Answer 1

What do you mean by timeslicing? The whole point of TCP is that data is verified. Data won't just go lost even if the process/thread isn't responding. The kernel will hold onto it until the connection times out.

I'd recommend just starting the Threads listening on different ports with a callback to the master handler (that launches the Threads), and then handling the messages there.

You don't need multiple processes.

Answer 2

There is absolutely no problem with your scenario. Threads waiting on I/O operations (like reading from files, sockets, waiting on blocking locks, etc.) are blocked automatically from scheduling and are put to sleep until an input is detected. Then, the operating system wakes up the sleeping threads so they can read the input coming from the sockets.