Distributed Debugging

TotalView provides a distributed architecture that supports many different operating environments, including:

• Remote programs running on a separate machine from TotalView.
   
• Multiprocess programs running on a multiprocessor machine.
   
• Multiprocess programs running on a cluster of homogeneous machines.
   
• Distributed programs running on a set of homogeneous machines.

  Note:   Distributed debugging requires that all machines have the same architecture and operating system.

The machine on which TotalView is running is known as the host machine, while the machine on which the process being debugged is running is the target machine. The host and target machines can be the same machine.

If the host and target machines are different, TotalView starts a process on each remote target machine. TotalView communicates with this process by using standard TCP/IP protocols.

TotalView Debugger Server

Debugging distributed programs does not differ from debugging nondistributed programs: TotalView offers the same set of features for each.

Depending on the platform, TotalView can debug programs that use the HPF, MPI, IBM Parallel Environment (PE), OpenMP, pthreads, and Parallel Virtual Machine (PVM) libraries.