Earlier this week, I attended the OpenDev conference in San Francisco, CA.
The conference was focused on the emerging “edge computing” use cases for the cloud. This is an area that is of particular interest, not just from the obvious applicability to my ‘day job’ at Verizon, but also from the fact that it opens up an interesting new set of opportunities for distributed computing applications.
The highlight(s) of the show were two keynotes by M. Satyanarayanan of CMU. Both sessions were video taped and I’m hoping that the videos will be made available soon.
His team is working on some real cool stuff, and he showed off some of their work. The one that I found most fascinating, which most completely illustrates the value of edge computing is the augmented reality application to playing table tennis (which they call ping pong, and I know that annoys a lot of people :))
It was great to hear a user perspective presented by Andrew Mitry of Walmart. With 11,000 stores and an enormous (2mm??) employees, their edge computing use-case truly represents the scale at which these systems will have to operate, and the benefits that they can bring to the enterprise.
The conference sessions were very interesting and some of my key takeaways were that:
- Edge Computing means different things to different people, because the term ‘Edge’ means different things to different applications. In some cases the edge device may be in a data center, in other cases in your houses, and in other cases on top of a lamp post at the end of your street.
- A common API in orchestrating applications across the entirety of the cloud is very important, but different technologies may be better suited to each location in the cloud. There was a lot of discussion of the value (or lack thereof) of having OpenStack at the edge, and whether it made sense for edge devices to be orchestrated by OpenStack (or not).
- I think an enormous amount of time was spent on debating whether or not OpenStack could be made to fit on a system with limited resources and I found this discussion to be rather tiring. After all, OpenStack runs fine on a little raspberry-pi and for a deployment where there will be relatively few OpenStack operations (instance, volume, security group creation, update, deletetion) the limited resources at the edge should be more than sufficient.
- There are different use-cases for edge-computing and NFV/VNF are not the only ones, and while they may be the early movers into this space, they may be unrepresentative of the larger market opportunity presented by the edge.
There is a lot of activity going on in the edge computing space and many of the things we’re doing at Verizon fall into that category. There were several sessions that showcased some of the things that we have been doing, AT&T had a couple of sessions describing their initiatives in the space as well.
There was a very interesting discussion of the edge computing use-cases and the etherpad for that session can be found here.
Some others who attended the session also posted summaries on their blogs. This one from Chris Dent provides a good summary.