You are (almost certainly) running out of stack. See threads   use threads ( 'stack_size' => 64*4096 };

Beyond that, using 500 threads where each does so little work every 60 seconds is a very memory intensive way to do things. Consider using the asynchronous 'syn' protocol of Net::Ping to efficiently ping your 500 hosts with minimal memory usage.

Try using asynchronous I/O rather than threads--i.e., fire off multiple pings, but rather than using a blocking read to get the response, you process responses asynchronously as they arrive. Using select on an array of sockets could do the trick, or maybe you'd prefer something like POE or IO::Lambda::Socket which handle much of the mechanics for you. (Note: I've used neither POE nor IO::Lambda::Socket, so I can't further comment on them.)

I agree -- you should group your pings using select rather than use a 1-1 thread-ping model. Your thread failure is most likely due to memory issues. By default, threads are allocated a 16MB stack. 100 threads would use more than 1.6GB of memory. If you are really enamored of the thread-per-ping model, and you are using 5.10, you can set the stack size to something reasonable, like 128K (you probably don't need much if all you are doing is a simple ping).

fnord

one process handling the 100 theads here..how to create the multiple process in linux and how to run the same program for each process... do you have any example please help me

We are not used socket concept here just we have used for Net::Ping module to ping the server.

Where does it say that you must use one thread per server?

We have to monitor the server for every 1 minite... if the server is "up" every 5 min and if the server is "down" every 1 min we have pinged, that's way we have opened one thread for one serevr running that thread as continuoesly