Ali's switch model is kind of network bridge specified in IEEE 802.1D with each port connected with a system(subnet). It can automatically discover the mac address of destination of the packet in some way. Yes, if Ali's model is implemented, it should work for cluster. However, I don't need to totally mimic the real world network. I just want N systems to communicate with each other in some way. For this purpose, it is ok to have mac addresses fixed and get them known ahead by the switch model. That is why I want to know how M5 sets the mac addresses. My switch model is much like EtherLink model, the bandwidth (rate) is a parameter used to create switch model, as it is used in EtherLink. I feel my switch model is more easily implemented.
Xin ---- Original message ---- >Date: Fri, 7 Dec 2007 14:54:25 -0500 (EST) >From: Nathan Binkert <[EMAIL PROTECTED]> >Subject: Re: [m5-users] issues on switch model for cluster >To: M5 users mailing list <m5-users@m5sim.org> > >Why do you feel that this model is necessary? Can you tune the bandwidth >of the individual links in your model? To me it just seems that you're >making your model overly complicated if you're making it deal with link >bandwidth. > >> You may treat the EtherSwitch as a general EtherLink. >> The EtherSwitch has N Interfaces and N*(N-1) Links, >> when N=2, it has 2 Interfaces and 2 Links, which is the >> same model as EtherLink. (EtherSwitch also has inner class >> Interface and Link, as shown in my previous email) >> I have tested my EtherSwitch model in M5 in the case N=2, >> it works well. >> >> Xin >> >> ---- Original message ---- >>> Date: Fri, 7 Dec 2007 00:28:57 -0500 (EST) >>> From: Nathan Binkert <[EMAIL PROTECTED]> >>> Subject: Re: [m5-users] issues on switch model for cluster >>> To: M5 users mailing list <m5-users@m5sim.org> >>> >>>> Actually, the EtherLink model can be completely replaced >>>> by the EtherSwitch model. >>> You can do that, but you probably shouldn't. The EtherLink itself is what >>> manages the bandwidth of the links. The switch however will have some >>> sort of internal bandwidth that is far higher than individual link. If >>> you don't have the links, you're going to get the bandwidth wrong unless >>> you more or less implement the link in the interface. >>> >>> Nate >>> _______________________________________________ >>> m5-users mailing list >>> m5-users@m5sim.org >>> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users >> _______________________________________________ >> m5-users mailing list >> m5-users@m5sim.org >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users >> >_______________________________________________ >m5-users mailing list >m5-users@m5sim.org >http://m5sim.org/cgi-bin/mailman/listinfo/m5-users _______________________________________________ m5-users mailing list m5-users@m5sim.org http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
