Minutes of Technical Community Discussion on May 12, 2016
Date and Time: 6:00am PDT / UTC 13:00, Thursday May 12, 2016
Convener: Bin Hu (AT&T)
- Artur Tyloch (Canonical)
Bryan Sullivan (AT&T)
Dan Druta (AT&T)
- Dave Neary (RedHat)
- David McBride (OPNFV)
Gerald Kunzmann (DoCoMo)
- Julien Zhang (ZTE)
- Mark Szczesniak (Casa)
- Maryam Tahhan (Intel)
Meenakshi Kaushik (Cisco)
Michael Lynch (Intel)
Stuart Mackie (Juniper)
- Weidong Shao (Huawei)
- Yuri Yuan (ZTE)
- Zhijiang Hu (ZTE)
Active Project Proposal Discussion
Bryan (AT&T) gave an brief overview of VES. It was aligned with an important component DCAE of AT&T's ECOMP. It is targeted for a common event data model. Details are shown in wiki VNF Event Stream.
Dave Neary (RedHat) asked if it fits all types of telemetry, such as infrastructure, hardware, or if it only applies to VNF telemetry. Bryan responded that convergence is the ultimate goal.
Dave further asked if the effort purely focuses on normalizing messages, or events. Bryan indicated both, i.e. the data model of event messages. It may leverage NETCONF/YANG, TOSCA, JSON, XML etc.
Dave also asked if there is plan to integrate with existing messaging system, e.g. RabbitMQ. Bryan responded that we don't want to over burning existing messaging system, thus it may not directly plug into messaging app. Dave said that it could be a different (instance) of messaging app. Bryan acknowledged it could be. More importantly, what technology to leverage is subject to discussion.
Maryam Tahhan (Intel) asked if Neutron Instrumentation is considered. Maryam then shared the link, and Bryan agreed to further look into it.
Maryam further asked how scaling is considered. She had a concern of adding another layer of overhead. Bryan answered that we will minimize the connection among VMs. There are 2 types of events to handle. One is machine-generic events, and the other is VNF/app-specific events. Bryan further explained that in ECOMP, there is an event collector (DCAE) sitting directly above infrastructure. There will be plugins for OpenStack so that those data can be consumed by OpenStack. Bryan agreed to add a diagram to further clarify it.
Maryam asked if any open source is considered. Bryan agreed to look into it.
Dave asked because of there are lots of interest here, not onlt VNF, but also infrastructure, data fast path, etc., Dave suggested to discuss use cases etc in mailing list. There is only one committer here. We need to look for more contributors. Bryan agreed.
Actions for Bryan:
- Look into Neutron Instrumentation for potential leverage,
- Add architecture diagram
- Look into open source solutions in this subject.
- Start discussion of use cases in mailing list to further clarify motivation and need, and encourage more participation/contribution.
Zhijiang (ZTE) suggest to postpone it to next week, because there is only 10 minutes left.