Did you turn away from your terminal and miss Infrastructure as a Service? Well, don’t worry, as there is a new paradigm for the techno-hungry audience – and for a change, this one actually has something new about it! Most technology trends are re-manufactured from the last trend that did not gain the traction that its primary sponsors intended. When I say “new,” of course it’s not “new new.” It’s a way of working that is enabled by the confluence of existing technologies that have been packaged in such a way that its use can be widely adopted. Sometimes, working in IT reminds me of a great line from the TV show “The Big Bang Theory” in which Penny asks, “So what’s new in the world of physics?” To which Leonard simply replies, “Nothing.” However, being totally new is no strong measure of the potential for disruption, and this has some real potential for disrupting the way apps and microservices consume the infrastructure foundation of the Internet.
I talk of Serverless Computing. The basic idea is that code/functions are developed and stored in the cloud in proximity to a related service. They are then triggered by events and APIs that execute the code using metered server resources, allocated, scheduled and maintained by the cloud provider. The resources have no persistent ownership or burden of management to the developer or organisation, and the consumer (code ‘owner’) is charged for the resources consumed and nothing else. The consumed resources are detailed to a granular level, and can be seen here for AWS Lambda, Microsoft Azure Functions and soon at Google Cloud Functions. Micro-pricing to construct micro-services. IBM have also made some moves here with their Bluemix OpenWhisk development – with source available on the ubiquitous GitHub.
“Doesn’t that sound like PaaS?” I hear you say – service APIs, triggers, etc.? Well yes, but the difference lies in the generic nature of the services that are available and the perception of limitless scalability. Add to that the pure ambition of the providers to enable their vast ecosystem of services, and it creates a stark contrast between the realities. PaaS has tended to lean towards the industries of least resistance, and when the coverage stretches wider, scalability constraints make you draw undesirable boxes around portions of your app. And while providing huge reductions in the ‘mean time to product,’ PaaS efficiencies are paid for through enforced residency in the ‘walled garden’ in which the APIs or platform operates. Therefore, what you gain in simplicity, you can lose in versatility.
When it comes to the construction of mega infrastructures, being prerequisite to the removal of scalability as a concern, it makes me think ‘some things only governments can do.’ In this version of reality, AWS, Microsoft and Google are the world superpowers: each have an oval office, each control a substantial arsenal of power and each consider the others as both competition and inspiration. The outward message is that a citizen is free to move between the providers. But of course, to take advantage of the ecosystem, you need to settle, customise and generate some technical debt, so movement is free as in puppy, not free as in beer.
Serverless can be considered an evolution of PaaS, depending on your viewpoint. For me, it is a different outlet. I think of PaaS as the ‘indie’ films of cloud – more specific audiences, lower budget, and a reliance on quality and quirkiness for adoption. Serverless is the blockbuster – released on the masses by the big studios, with merchandise in all the shops and a sequel already in the making.
Serverless Computing is a name that architecturally resonates on the negative. Even Werner Volgels (CTO of Amazon) in recent tweets has professed his dislike of the term, as it articulates what it is taking from the party, not what it is bringing. Personally I like the term DLIC (dynamic linking of infrastructure as code) as it describes the positive intent at the foundations of the architecture. But even this is not perfect, as to some degree it’s still giving a respectful nod to a changing industry. It may be time to think about putting servers in the same category as industrial transformers – in the 1880’s they were the bleeding edge of electrical transmission, allowing the connection of cities and creation of power grids. But now, while we still need them even more than we ever have, they are grey functional items that have left the limelight and just work. The server is dead, long live the server.
Feedback is always welcome, so feel free to get in touch at @glennaugustus
comments powered by Disqus