Thanks for your reply. It is very helpful!
On Tue, Sep 24, 2013 at 5:53 PM, Wes Felter <w...@felter.org> wrote: > On 9/24/13 3:26 PM, Ming-Chen Zhao wrote: > > From these old research papers, I can see the flow setup rate they can >> achieve is about 200-300 flows/sec. The bottleneck is the bandwidth >> between controller and openflow switches. >> > > We saw more like 1,000 flows/s on the G8264, so things have improved a > little. > > > However, there are no reports >> from industrial to discuss what the real flow setup rate they need... >> Can 200-300 flows/sec be enough to support a small datacenter network? >> > > One way to begin to answer this question might be to consider STP or OSPF > convergence time and compare it against OpenFlow. In some cases RSTP can > converge in less than one second, but OpenFlow convergence time is > proportional to the number of flows; at L2 the number of flows is at least > the number of hosts. So an L2 OpenFlow network will converge slower than > RSTP if it has a significant number of hosts. > In an L3 network the number of flows might be equal to the number of > subnets, so a small number of updates per second might be acceptable. > > > -- > Wes Felter > IBM Research - Austin > > ______________________________**_________________ > openflow-discuss mailing list > openflow-discuss@lists.**stanford.edu<openflow-discuss@lists.stanford.edu> > https://mailman.stanford.edu/**mailman/listinfo/openflow-**discuss<https://mailman.stanford.edu/mailman/listinfo/openflow-discuss> > -- Mingchen Zhao Ph.D. Candidate, Computer and Information Science Program (CIS) School of Engineering and Applied Science (SEAS) E-mail: miz...@seas.upenn.edu
_______________________________________________ openflow-discuss mailing list openflow-discuss@lists.stanford.edu https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
