Silicon Valley is a unique place – you have an idea or prototype, get it out, and very early you will get feedback about what`s the appetite for it. I have been running meetups in the valley for last 3 years around Cloud Foundry, OpenStack and Bluemix, and have always felt meetups are the best mechanism to feel the pulse, vibe, and get some early honest¬†feedback about technologies.
Very recently we at IBM open sourced a project around serverless technology, OpenWhisk. And last week I ran the first full fledged meetup on it, partnering with Altoros, community partners who have been also working on OpenWhisk. Andrei Yurkevich, Altoros CTO was the co-presenter.
This was a follow on to the lightening talk I gave on OpenWhisk during the CFCamp at CFSummit. I talked about how OpenWhisk can be the perfect platform to build cloud native serverless microservices using containers, which are triggered by external events. In fact for most of the 12 factor app commandments, OpenWhisk seems like a perfect fit.
— IBM Bluemix¬ģ (@IBMBluemix) May 23, 2016
I want to note some of the observations as we introduced OpenWhisk to the valley audience, and answer some of the questions we received.
What does Serverless mean? No servers? Isn’t that what we get by consuming any public cloud services?.
It means that your code is not consuming any sever cycles while sitting idle. Essentially in traditional models, you still have the code running, whether you are serving requests or not. Another way to think of it is framework for event driven microservices.
We would still need servers for any persistent storage, like database, object store etc?
Yes. And that`s also true for any cloud APIs you will be calling – they will be running on their servers. This gives you a way to write your application logic in a way so that its only executed in response to events, and you are paying for that duration of action invocation.
Where does it leave the Cloud Foundry PaaS or Containers landscape – is this going to be replace them?
My¬†point of view is no, this is not an alternative to them. Its another way of consuming cloud platform. If you have long running processes and applications, Cloud Foundry still should be your first choice. I am a big fan of Cloud Foundry myself, and its the best platform in my opinion for long running cloud native apps. If you want to control the operating system, VMs should be where you need to focus on, and if portability is your first choice, containers remain the answer. My point of view – VMs and containers in the end are means to a larger goal – an application platform, and that`s where Cloud Foundry and OpenWhisk fit.
What`s the main differentiation you have with Lambda, Azure functions etc?
In short, OpenWhisk is¬†the first open source event driven serverless framework available, and the first one which you can install on your hardware. There are other distinctions, specifically focussed around our cognitive functions etc.
You described the concept of cold and warm containers in queue – wont warm containers consume server capacity
They are kept for a very short configurable period of time, and in a paused stage.
Is there any predictive analysis mechanism which feeds into when actions will more likely be executed?
Not currently. This is a reactive model
Overall it was a great evening, and I enjoyed the interactions¬†and some very insightful questions. Even though its early days, its great to see a positive vibe towards this technology in Silicon Valley, and hopefully very soon it would see huge¬†community participation and widespread adoption.
— Animesh Singh (@AnimeshSingh) June 15, 2016
Very soon, we are planning to to go in further technical details at multiple conferences, so if you are interested in the technology, please keep watching my twitter handle @AnimeshSingh for further conferences and meetups.
— @ThingsEXPO (@ThingsExpo) June 20, 2016