#opnfv-meeting: LSOAPI
Meeting started by KLuehrs at 16:01:22 UTC
(full logs).
Meeting summary
- Welcome (KLuehrs, 16:01:40)
- Randy Levensalor (RandyLevensalor,
16:03:38)
- Extension of project scope: MEF Class of Service and Bandwidth Profile requirements (KLuehrs, 16:03:58)
- Steve S analyzed MEF Class of Service
requirements and Bandwidth Profile requirements and proposed a
limited set of attributes to define scope. (KLuehrs,
16:07:38)
- The core model needs to provide sufficient
functionality while not over-scoping, in order to make rapid
progress. (KLuehrs,
16:08:33)
- Mehmet reinforced the importance of maintaining
consistency with MEF specifications. (KLuehrs,
16:09:18)
- LSOAPIs should be consistent with APIs
developed by Metro Ethernet Forum Service Activation and
Configuration (SCA) project (KLuehrs,
16:11:33)
- Steve S shared presentation 'LSOAPI Phase 2 -
Scoping Considerations (Sept 23 2015) (KLuehrs,
16:13:29)
- Proposed: include MEF ELine/EVPL single
operator and E-LAN single operator service in scope (KLuehrs,
16:14:24)
- Proposed: Out of scope service for now: E-Tree
and Multi operator services (KLuehrs,
16:15:04)
- Mehmet agrees limiting scope to single operator
service. Focus on ELine initially for this phase. (KLuehrs,
16:16:04)
- Mehmet does not have a strong preference on
which ELine service (EPL or EVPL) to focus on first (KLuehrs,
16:18:58)
- Recommendation for scope of LSOAPI, for the
Brahmaputra release timeframe, is limit to ELine service.
Additionally some research is needed to understand how to
merge/incorporate/interface with work done to define APIs defined by
MEF SCA project to configure network elements between UNIs.
(KLuehrs,
16:23:39)
- Proposed: Include in scope VLAN tagging
support (KLuehrs,
16:24:02)
- Proposed: LAN packet priority tagging (PCP,
DSCP) out of scope (KLuehrs,
16:25:30)
- Proposed: CoS and Bandwidth profile per PCP /
DSCP out of scope (KLuehrs,
16:25:58)
- John C responded to say it's important to
preserve priority tagging (KLuehrs,
16:26:40)
- Mehmet asked if this means we will not be able
to provide Class of Service (KLuehrs,
16:27:01)
- Steve S responded it gets down to how much
detail we include to provide CoS. MEF supports providing CoS to EVC
as a whole (all traffic in EVC subjected to same CoS). Another
option defined by MEF is application of different CoS within EVC
based on Priority Tag. Scoping question for APIs is how granular we
need to support priorities. (KLuehrs,
16:29:18)
- Steve S said if we don't support Priority Codes
we can still support CoS per EVC. (KLuehrs,
16:29:52)
- John C - Charter Cable - Moving from single CoS
per EVC to multiple CoS per EVC. Working toward CE 2.0 compliance.
Want to be sure we do not preclude including in the information
model and APIs multiple CoS per EVC (priority tag). (KLuehrs,
16:31:47)
- Mehmet: Agree CoS per EVC to start with
(KLuehrs,
16:33:29)
- Bill Armstrong - Charter : couldn't we include
all the priority codes in the API and model and simply map them to
EVC, and use them as we need to later. (KLuehrs,
16:35:28)
- Steve S - It's not difficult to include the
priority tags in the model. It sounds like there is interest among
operators to include the priority tags. (KLuehrs,
16:36:25)
- Steve S reviewed CoS requirements in MEF 10.3
specification: nine Class of Service attributes (KLuehrs,
16:38:12)
- For any of the CoS metrics, e.g., one-way frame
delay, a set of performance parameters is defined. (KLuehrs,
16:38:58)
- Do service providers need CoS Performance
Parameters included in the API, or is it sufficient to include a
simple value for each CoS metric? (KLuehrs,
16:42:49)
- Is it possible to define each CoS metric as a
container, so that we can include a single 'aggregate' value for
now, but potentially extend the table later? (KLuehrs,
16:43:47)
- Steve S: Choices include (1) create first
container with single metric only for now, (2) create container for
each metric and include all the performance parameters, ...
(KLuehrs,
16:45:52)
- Ioakem S - How will the CoS attributes be
validated? How will we know if API is implemented correctly? We will
need a test bed with particular devices. (KLuehrs,
16:48:28)
- Mehmet - Referring here to MEF certification.
Certification test cases will validate adherence to the MEF
specification (KLuehrs,
16:49:37)
- LSOAPI project's job is to provide the means
for the correct parameters to be passed. It will be the job of the
application implementers to ensure the parameters are used
appropriately. (KLuehrs,
16:50:46)
- Bill A - Don't see any reason why we can't
model the CoS metrics and performance parameters. As long as they
parameters are just pluggable values and not dependencies it should
be reasonable to model them. (KLuehrs,
16:53:56)
- Steve S - next MEF parameter group is Bandwidth
Profiles (KLuehrs,
16:54:26)
- Bandwidth Profiles define how to handle traffic
entering and exiting the service provider's network (KLuehrs,
16:55:05)
- MEF specifications address Ingress Bandwidth
Profiles in significant detail and spend less time on Egress
Bandwidth Profiles (KLuehrs,
16:55:50)
- In MEF specifications Egress BP are optional.
Should they be in scope? (KLuehrs,
16:56:31)
- John C - We should include Egress BP into scope
if multiple profiles are to be supported. (KLuehrs,
16:57:51)
- Team participating in today's call agree it
makes sense to keep multiple profile per EVC out of scope for
Brahmaputra release (KLuehrs,
16:58:36)
- Steve S - Next MEF specification topic is Token
Sharing. Steve provided brief explanation of token sharing.
Single-flow processing is defined in MEF as a way to not need token
sharing. (KLuehrs,
17:00:01)
- Proposed: Token Sharing out of scope for
now (KLuehrs,
17:00:24)
- Bill A - this does not become an issue until we
go to multiple CoS (per EVC) (KLuehrs,
17:00:55)
- Does this apply to multiple UNIs per EVC?
Refers to borrowing bandwidth between EVCs. (KLuehrs,
17:02:01)
- This is a complex problem to solve so can be
deferred. (KLuehrs,
17:02:30)
- Continue EIR discussion next week (KLuehrs,
17:03:24)
Meeting ended at 17:03:27 UTC
(full logs).
Action items
- (none)
People present (lines said)
- KLuehrs (47)
- collabot (4)
- RandyLevensalor (1)
Generated by MeetBot 0.1.4.