All the IT excitement at the moment seems to be focused on the concept of “serverless” computing. In particular, Lambda from AWS and Open Whisk from IBM are generating huge amounts of interest in the idea that, finally, application developers can concentrate on building code without having to factor in what servers and operating systems it will run on.
Serverless is somewhat of a misnomer. Ultimately, the apps and micro-services developed in this way will run on servers somewhere. Currently, that will be in the data centres of large public cloud providers. Ultimately, however they could run in any private or hybrid-cloud infrastructure as well. It is just that the developers don’t need to take any account of this. Even in current cloud deployments they need to configure DNS entries, specify VM sizes and perform a host of other tasks necessary for the deployment and maintenance of servers.
This is not, yet, the silver bullet that removes all the complexities and decisions about how and where to deploy all your applications.
The key feature of serverless that makes it excellent for some applications and not for others is its ability to only call on server resources when they are required. This makes it particularly suitable for applications that are called on at specific times and then only require a short burst of processing. The sweet spot is the Internet of Things (IoT). Sensors collecting information may need, periodically, to trigger a business process or function. Only when that function is called upon does a service on a server, somewhere, come to life. As soon as the function has been processed, the service shuts down again. For a great example of how this works for rubbish collection have a look at this article by David Norfolk at Bloor Research.
These sorts of functions can be delivered much more cost effectively using a serverless architecture.
We are already starting to see Function-as-a-Service offerings that enable you to assemble applications quickly from an array of pre-existing micro-services. But it has limitations around configuration, security, latency and data retention that don’t make it ideal in many use cases. This article by Mike Roberts is a great primer on serverless architecture. Don’t worry if you are not particularly technical, scroll down to his section on benefits and drawbacks and you’ll get a clear picture of where this might, or might not work for you.
Behind this shiny new tech story is a broader, more fundamental issue, that we have discussed before, but is worth re-stating. It is the applications and functions you need to run and develop your business, not the underlying IT infrastructure, that are the starting point in any IT strategy. Even when you have worked out which applications can be moved to the Cloud, there are, as we can now see, different ways they can be deployed. What architecture, methodology or IT tools set needs to be used will be dependent on the applications.
We believe in a strong, data driven methodology for driving transformation.
It has helped us work with the Boards of numerous organisations, large and small, to manage a journey to the Cloud that is business function driven, not technology led. Let us know what experiences you have had in trying to make that journey. We love your stories. They help us shape and improve the services we deliver.