[edge-wg] Eclipse Edge Native WG call minutes — September 16, 2020

[Frédéric Desbiens <frederic.desbiens@xxxxxxxxxxxxxxxxxxxxxx>]

Hi everyone.

You will find the minutes from the September 16, 2020 call below. Minutes from prior meetings are also available in this document:
https://docs.google.com/document/d/1XIv3-SYK-HadgScJ_H4tKvwRhGsig7uLg7n0Xzvwza0  

Meeting recording

https://eclipse.zoom.us/rec/share/1UvdRF5obCnC7V2eklQ5aOPd0vm99FaHsJj5Fyrx-axq6RoD5FpCxzsoqmxkPTt0.ecE7eGlNW_iNSj8i Password: 1O@!!EWr  

The next call will be held on September 16, 2020.

=====

1. Sensors at the Edge: How can Edge infrastructure help mitigate sampling rates that are excessive given the business requirements.
   
   Barry Kaplan started the discussion by explaining what he did and what he intends to do to evolve his company’s product, which is about 10 years old.
    

• Current situation

• Collecting data from various edge devices (factories, connected vehicles). Not doing a lot of processing at the edge right now; but trend is to push more of the processing to the edge.

• Data is collected in bulk and not filtered or simplified. Published in batches to an MQTT gateway in the Cloud. Uses Artemis as a gateway to Kafka.QoS acknowledged by the gateway.

• Raw data forked to several destinations. 1) Stored as is in long term block storage (S3, GCP) for future reuse and reprocessing. 2) Written to a time series DB (historian) such as Elastic / Influx. for other systems to process. 3) Steam processing (real-time), to look at patterns in the data and generate alerts. 

• Looking at how they could leverage Hono. Maybe replacing MQTT with AMQP. 

• Security and device registration are a challenge.
  
  

• Move to the edge

• Need to deploy microservices at the Edge. But how do you manage edge nodes?

• Running k8s at the Edge is attractive. However, k8s will not fit on small edge nodes. ioFog attractive in those scenarios.

• New architecture: data is routed at the Edge but still flows to the Cloud. Could start filtering the data at the Edge because a lot of the data transmitted is not used.

• Would be nice for Hono to be directly integrated with Kafka, bypassing AMQP transport.

Barry’s introduction started an exchange which lead to the next agenda item.

2. zenoh overview
   Luca Cominardi delivered a presentation on zenoh, the pub/sub protocol at the core of Eclipse fog05. zenoh addresses many of the challenges mentioned earlier in the discussion.
   
   
   Additional resources:

• Code repository: https://github.com/eclipse-zenoh/zenoh (use the rust-master branch)

• Zenoh mailing list: https://accounts.eclipse.org/mailing-list/zenoh-dev

• Zenoh website: http://zenoh.io/ 

• Video presentation: https://www.youtube.com/watch?v=B0WVHnwf7es 

• Gitter channel (for any information, questions on Zenoh): https://gitter.im/atolab/zenoh
  
   

3. Edge Data encryption (data at rest)

There was no time to cover this point; it will be added to the agenda of the next meeting.

=====

Best Regards,

Frédéric DESBIENS

Program Manager, IoT and Edge Computing | Eclipse Foundation, Inc.

Twitter: @BlueberryCoder

Join us at our virtual event: EclipseCon 2020 - October 20-22