Notes from functions17 conf in Toronto

IBM cloud has support for open source apache open whisk. which aims to be could provider independent serverless framework. Will keep an eye on their progress.

1. Google’s firebase functions are in beta and hence by design do not provide any SLA for any serious work.
2. Azure does not have mac support yet. The workflow is bundled with VSCode IDE and the whole microsoft developer ecosystem. However during the talk I was able to set up a hello world kind of simple function. (Hodor as a service) from their in browser-editor. Doesnt feel like production ready, more of a demoable small stuff. THey have another service called Logic App, which sounded like they are trying to emulate something like zapier and integrate it with custom functions.
3. AWS had by far the most mature understanding of functions and their customer needs. Unsurprising. Talk from their side was not really a demo or small tutorial of their system but how 12 factor app is still relevant in the serverless world (Except 4.5 out of 12 points) and by designing functions properly the guidelines can help designing the system.
4. Pivotal folks were demoing a bleeding edge of their open source work on “bosh”. They are doing some kickass work on abstracting out cloud providers so you can switch them easily for running serverless. Although their demo of java/spring functions went above my head.
5. Overall, it was very evident that most work in serverless world is done by node folks and they have the best tooling so far. To actually use the cutting edge functions infra or tools, I may have to drop pip install for a while and do a npm install.
6. One of the interesting talks was by the Keynote Kassandra Perch from IoPipe, and her experience of using lambda with longboard. She had set up a contraption with her longboard that when it crashes, her pendent would shine or something and she has to press a button in 10 seconds else her room mate will get a panic email (through a serverless function hosted on AWS lambda). iopipe seems to be working on interesting problem of application monitoring, tracing on lambda functions in production. They recently launched a feature where you can see how much time your code took line by line. Can be interesting to see how this works because in my opinion this should not be needed. All the business code and actual work should be abstracted out in the external libraries. Actual aws lambda code just need to be like a view in MVC.. very thin. IN any case building tools for serverless functions is hard because the vendors are in a flux, so kudos to them.