--
wg-SDN call notes, 11/20/14
Attending:
Dan Schmiedt, willys@clemson.edu
Michael Lambert <lambert@psc.edu>
Michael Van Norman, mvn@ucla.edu
kevin mayeshiro, kmayeshiro@ucdavis.edu
Rich Cropp, rac111@psu.edu
Dave.Pokorney@flrnet.org
Charles Chambers, chambers@uh.edu
AJ Ragusa (aragusa@grnoc.iu.edu)
Kathy Benninger benninger@psc.edu
Dale Finkelson not on Etherpad
Heidi Dempsey, hpd@bbn.com
Agenda:
0) Agenda Bash
1) AL2S update, Internet2/GRNOC
AJ: in final stages of OESS 1.1.5 testing.
M. Lambert: Will there be a "logout" button in the GUI?
AJ: Issue is Shiboleth: not a standard way of doing it at all sites.
M. Van Norman noted that there is a "logout" function in Shib. AJ will investigate, but noted that shin isn't the only auth method.
AJ noted that they are beginning to develop a new branch of OESS that uses the Ryu controller in order to begin working with/testing OF 1.3.
Bill Chimiak noted that Ryu seems to be tilted towards production networking and that in general this seems like "a good thing."
AJ noted that it only took ~1.5 days to port initially to the Ryu controller and that a good API mad sit possible really abstract away the controller.
M. Lambert asked if this required OF 1.3-capable switches. It does not.
Primary testing is with Juniper and Brocade switches, as deployed on the I2 network.
2) SDNTrace update: Joe Breen & Deniz are traveling today and likely won’t make the call. Initial proposal will be sent to the wg-SDN listserv in the next week.
Dan S. noted that Joe Breen and Deniz Gurkan are working on an NSF-style proposal in order to spec out the initial SDNTrace. This proposal will be shared with and discussed on the wg-SDN listserv and on future calls.
AJ noted that there is a ticket to add traceroute functionality in OESS 1.1.5. He has discussed this with Joe Breen at SC14. AJ and the wg-SDN will work together in order to synchronize/minimize wheel reinvention.
3) Bill Chimiak asked if there were any issues operating a 100G port on a Brocade MLXe in Hybrid mode. AJ noted that the only issue is that L2 is punted to the CPU when the port is in Hybrid mode. So, >1G traffic passed via VLAN at L2 would likely be CPU-bound. Heidi Dempsey noted that similar issues had been experienced in GENI, but not at line rates that high.
4) AJ asked the group to study Scipass: http://globalnoc.iu.edu/sdn/scipass.html which is an OF-based load balancer for security appliances, and make suggestions on improvements/usefulness.
End of call at 2pm EST