Hi,
Hello all!
Perhaps it does not make sense this way, but I thought about reducing the node count for some congestion scenarios I'm currently working on.
By running 2 separate instances of GIST and NSLP code, each pair correctly doing their socket-API calls, and having 2 NICs on the PC, can I fake multiple nodes? (For example, in RMD, 1 egress and 1 core running on the same pc, associated with one NIC and IP address.)
Can you give me a reason, why you want this, but not a virtual machine? I mean, in terms of any performance study, this kind of setup is not realistic anyway. I just ask out of curiosity.
Is there a working solution for running multiple (for ex. 2) GIST instances on the same machine, to which are attached separate NSLP instances correctly?
No, there are several problems. The first one is the API to the NSLP as you noticed. NSLP and GIST communicate through a unix domain socket which has a fixed name, so the two running instances would collide. This can easily be fixed by compiling a second instance of NSIS which uses a different name, of course. The second issue is more complicated. It is about sockets. GIST needs to intercept packets with router alert option. For that, we open up a raw socket, that monitors any device for packets with RAO. You would have to chance this and I currently do not know how to do this, so I am not able to help with that. There might be more problems, but no one that directly comes into my mind.
Sorry, that I can not provide more help.
Christian Dickmann
nsis_imp@informatik.uni-goettingen.de