#opnfv-sfc: SFC F2F at ODL Hackfest
Meeting started by bryan_att at 16:36:15 UTC
(full logs).
Meeting summary
-
- Carlos Goncalves (cgoncalves,
16:39:13)
- Bryan Sullivan (bryan_att,
16:41:29)
- http://eavesdrop.openstack.org/#Neutron_Service_Chaining_meeting
(cgoncalves,
16:45:03)
- https://git.openstack.org/cgit/openstack/networking-sfc
(cgoncalves,
16:47:20)
- A project has been initiated in
OpenStack (bryan_att,
16:48:49)
- ... Brady provides an intro into the OPNFV SFC
project so far (bryan_att,
16:52:33)
- https://docs.google.com/presentation/d/1JOWN-Xf0_GrV5v_n5q9_j9nUQcM8MjlNl3Sw1YN1bxU/edit?usp=sharing
(bryan_att,
16:55:03)
- Slides presented at ODL Summit (bryan_att,
16:55:17)
- Brady directs us to slide 5 of the OPNFV SFC
working deck (bryan_att,
17:07:02)
- https://docs.google.com/presentation/d/1gbhAnrTYbLCrNMhMXin0lxjyg-7IHNPjrlBTIjwAzys/edit#slide=id.gb69fe6cbf_1_0
(bryan_att,
17:07:15)
- Are SFs assumed to be NSH-aware? (bryan_att,
17:07:57)
- Maybe for some... there are variety of issues
to be addressed if they are not, e.g. scalability (bryan_att,
17:08:37)
- OVS support for NSH is another
dependency (bryan_att,
17:09:01)
- There is a patch pending for this in OVS, and
we have to have official release with this patch (bryan_att,
17:09:31)
- The IETF draft Rev 1 should be published soon,
mostly a cleanup of Rev 0 - no change to arch etc (bryan_att,
17:10:26)
- DPDK-based OVS is being pursued by Intel, and
is part of OVS mainstream, and is a patch Intel controls
(bryan_att,
17:11:29)
- what version of ODL will B release be using?
Lithium is assumed, but there may be some features we need from
Beryllium (bryan_att,
17:12:07)
- What OVS version do we need? 2.4 is late Aug,
2.5 will include Contract (bryan_att,
17:12:44)
- If its in the official release then ok but a
private branch is not (bryan_att,
17:13:37)
- clarifying the point above - the VM will need
to be NSH-aware (specifically the kernel and/or OVS) rather than the
application (SF) running in that VM (bryan_att,
17:16:44)
- Brady: next issue being considered was how VMs
are launched as part of SFC setup (bryan_att,
17:17:37)
- Uri: asks about the role of the "SFC-Openstack
coordinator" (bryan_att,
17:18:13)
- Brady: what's the trigger to start the VMs is
the focus of the coordinator (bryan_att,
17:18:43)
- ... Some discussion about starting a new
project in ODL related to this (bryan_att,
17:20:28)
- ... slide 6-8 are 3 options for how this could
work (bryan_att,
17:20:42)
- ... we agreed to go with option 2 (slide 7) .
In this slide the VNFM is the "coordinator" from the earlier
slide (bryan_att,
17:22:06)
- Paul: The user doesn't care just needs a FW,
LB, etc (bryan_att,
17:23:15)
- ... e.g. log into AWS and ask for the function,
you don't specify the type (bryan_att,
17:23:53)
- Brady: we will need the option for both
(bryan_att,
17:24:18)
- https://docs.google.com/presentation/d/1gbhAnrTYbLCrNMhMXin0lxjyg-7IHNPjrlBTIjwAzys/edit#slide=id.g5bc8f02e2_53_0
(paulq,
17:24:43)
- Reinaldo: need to show how to go from slide 7
to a chain (slide is about setting up a SF) (bryan_att,
17:25:06)
- ... the whole problem is who is the master of
the VNF? (bryan_att,
17:26:09)
- Keith: something needs to coordinate the
compute, storage, network elements; we need to consider the series
of events and the level of abstraction (bryan_att,
17:29:02)
- Chris: the orchestration part is more clear to
us, our focus is on the MANO stack (bryan_att,
17:29:46)
- Bryan: we also have a lot of interest in that
function, monitoring inventory and allocating resources at the
VM/VNF and service (chain) levels (bryan_att,
17:31:01)
- Brady: there's an error on slide 9, originally
thinking SFFs would live in their own bridges in OVS (bryan_att,
17:31:46)
- Keith: yes the SFFs will live in br_int along
with GBP otherwise NSH info will be lost in the transfer (it wont
work) (bryan_att,
17:33:54)
- Brady: this is where we have gotten to in the
meetings - now that we have option 2 as our plan we will dive into
the next level (bryan_att,
17:34:38)
- ... e.g. the VNFM and the initial installation
of the VMs - Dan Smith will help with Fuel installation (bryan_att,
17:36:16)
- Bryan: we should setup a demo in parallel on
multiple implementations if possible e.g. the APEX and/or JOID based
installs if we can get support (bryan_att,
17:37:27)
- Bryan: I will refer to SFC and GBP in the
Copper project as the mechanisms to make service chains work, and
focus on the overall VNFM/NFVO functions to coordinate (bryan_att,
17:42:13)
- The ODL API browser provides the details on the
API (bryan_att,
17:42:54)
- the GBP-SFC demo is available on github, and
you can see the API docs thru the UI (bryan_att,
17:45:08)
- https://github.com/alagalah/gbpsfc-env
(bryan_att,
17:45:11)
- how will we deliver the VNFM etc parts, e.g.
where (what upstream project) and based upon what needs to be
determined (bryan_att,
17:49:01)
- Brady: we may have to carry some glue code in
the OPNFV-SFC code base for the time being (bryan_att,
17:49:32)
- Bryan: we should try to avoid carrying code
over a release boundary (bryan_att,
17:49:53)
- the options include XOS, Tacker, OpenMano,
etc (bryan_att,
17:50:15)
- OpenStack contributors are trying to get SFC
into it but progress is unclear (bryan_att,
17:50:45)
- Uri: we can carry a test VNFM driver for the
short term, knowing that it is not the final solution and that might
give us some time to work out the upstream (bryan_att,
17:52:05)
- Bryan suggests that we include some
documentation on the stack elements underlying SFC using NSH
(bryan_att,
17:58:15)
Meeting ended at 17:59:39 UTC
(full logs).
Action items
- (none)
People present (lines said)
- bryan_att (49)
- cgoncalves (3)
- collabot (3)
- alagalah (1)
- paulq (1)
Generated by MeetBot 0.1.4.