[cgl_discussion] PoC concall minutes, 2/27/03

Rod Van Meter Rod.VanMeter at nokia.com
Thu Feb 27 09:07:41 PST 2003


Ericsson: Jon Maloy
Nokia: rdv
OSDL: Randy, Andy, Mark, John, Craig
HP: Khalid
Intel: Julie, Rusty

rdv's trip report on Japan -- not a lot relevant to PoC was decided.
Good first set of meetings, though.  Will have to follow up with them
to get more firm commitments.  People in CGL are curious what DCL work
the Japanese companies are doing.

Load TIPC module, it starts broadcasting ethernet packets (or UDP
multicast) letting others know of its existence.  Gives you very
dynamic membership.

Subscribe for the existence of new processors or processes, and you'll
be notified, you can register with a table.

Using ethernet, not limited to a single cluster per ethernet.  Triple
<Zone id, subnet id, node id> defines a cluster.  All having the same
zone id form a cluster.  Can't have more than 16 clusters in a LAN, up
to 512 nodes per cluster.

Support inter-zone links.

Jon Maloy thinks up to a thousand nodes will probably work; 52 is the
most they've tested.

John Cherry says OSDL has three nodes, Mika is playing with it.

JonM would like feedback on the code; he has a long to-do list in the
README in 0.93.  He would also like to see a comprehensive test suite
for this, since their test SW isn't appropriate.

Rusty is hoping to get Intel heads in China to work on the test suite.

According to Stephanie, LTP test rig can do network tests.

John will take a look at TIPC socket notification APIs to compare to
OCF stuff.

Randy on SNMP: Yixiong working net-snmp side, Randy working on
kernel.  At first blush, there's a fair amount of work to be done in
the kernel.

RFC2465 has six tables; one is pretty straightforward (just adding
counters), but the others are going to be quite a bit of work.  Randy
wants to consult w/ the community about the value and approach on
this.

Craig is still concerned about being stuck waiting for the TAHI guys
to release stuff.  I (rdv) will contact TAHI folks about this.






More information about the cgl_discussion mailing list