Serverless computing Q&A with Mitesh Desai

The Q&A

What is Serverless Computing?

Serverless computing is an architecture where code execution is fully managed by a cloud provider, instead of the traditional method of developing applications and deploying them on servers.

There are multiple benefits to Serverless computing. It takes away the burden of infrastructure management and admin, providing a scalable platform so users can deploy code quickly and efficiently. The increased efficiency means organisations can save money and reallocate resources to accelerate the pace of innovation.

How does Serverless computing save money for businesses operating in the cloud? Could you give an example?

From a commercial perspective, it’s far cheaper to run containers and serverless objects in any cloud platform, especially Azure, than it is to run virtual machines. So just as an example, a typical virtual machine that could be anywhere up to £300 could get replaced with a container that starts at £60-£70. The big cost difference there is based on the resources that you’re using in the cloud environment just to run that particular function or that code, whereas with a virtual machine, you always have resources that need to be allocated and turned on for the virtual machine itself to be present.

On average, we see around a 40 percent decrease in costs by moving from traditional virtual machines of different sizes. So this could be machines that are doing basic Web application type roles, to complex database roles. If you take a mixture of these two machines and move them into containers, plus some platform-as-a-service like SQL containers, on average, we see at least a 40 percent reduction in the costs.

OK, so that’s considerable, is it possible to save anything around licensing as well?

Depending on where how far you want to take your Serverless journey, it will give you a lot more cost savings. You could go to functions, which are pure containers that just run the framework for the code. So for example, if you’re looking at a Java website or application, those will completely reduce any operating costs because there’s no Windows or Linux costs involved. If you go to Kubernetes and Docker containers, you would look at going to either a free Linux version which again will reduce your cost or a RedHat version and potentially a Windows Docker image as well, all of those there are reduced costs compared to maintaining and running a full-blown operating system on a virtual machine.

Are there any situations where it wouldn’t make sense for me to move?

Sometimes it depends on what type of workload you’re looking to move. At Kocho we do a lot of pre-sales and workshops prior to looking at a Serverless environment. In some cases there are going to be legacy applications or applications that have been developed that do rely on a particular function, that a virtual machine or a full-blown operating system provides. This would mean that you can’t lift and shift the application into a Serverless environment. However, that’s where we can look at refactoring the application and modernising it to try to make it work in a Serverless environment.

If my business has automated configuration, will I have to re-do all that work? What’s the cost of moving to Serverless competing in approximate days if I have an environment, say, of 100 servers?
So typically we use Azure dev-ops. Most cloud platforms have a dev-ops program behind the transformation to this Serverless world, so that helps accelerate the journey and then helps you to scale going forward as well. So we look at typically anywhere up to 45 days to do a transformation of that type of workload into containers, whether they’re docker images, whether they’re Kubernetes or full Serverless in terms of functions and event groups.

It’s all driven through Azure dev-ops or another dev-ops program. So going forward to scale that environment or to shrink the environment down is fully automated and controlled through pipelines.

From a support perspective, how can I take advantage of new ways of supporting my environment?

Going forward, you’ll be using a lightweight O/S and Docker based containers and frameworks. That will help reduce your patching cycles, there will be less frequent patching of operating systems, less vulnerabilities to manage. Also just the general maintenance and management of the operating system will be easier. So from a technical perspective, you get those advantages.

From a commercial perspective, again, we’re seeing where we’ve done these transformations, there’s also a decrease in the managed service costs because there’s less patching for us to do as an MSP, there’s less maintenance work to do on servers. So that means further savings.

Are you seeing many organisations considering moving to Serverless computing now? It reminds me of the days when we started talking about going to virtual machines in Azure and everyone was quite sceptical. What’s the feeling you’re getting from the clients that you engage with?

I think it’s fairly opposite to that sort of scenario now, because I think a lot of the clients are realising that if they modernise the application, they’re not going to just benefit from re-platforming. So i.e. moving it from Virtual Machines to Serverless, but they’re taking advantage of scalabilities and actually addressing issues and problems that they have currently with their applications. Whether it’s a scale for performance or for more clients. So we’ve seen a lot of clients sort of say, well, we don’t want to move to this environment just because it’s cheap, or because it’s something new out there

A major reason to move is to enhance that resistance capability. So whether it’s to create lots of small environments for different clients to use their products, or whether it’s to shrink and rapidly increase that environment based on demand. So we’re getting a lot of traction from that, rather than a ‘let’s move to a different way of hosting our application’, because what we’re really saying is let’s modernise the application so it can fit with the ongoing business requirements.

Who is the market leader from a public cloud perspective? Because obviously we still have the the big players in that space, so Microsoft, Google, IBM, VMware through IBM. And I’ve noticed that we’ve got organizations like Rackspace offering their own flavor of Serverless compute…

So there are some strategic market players like Rackspace who are starting to offer their own container dockerised options. A lot of those are focusing around services that they already provide. Rackspace provides a lot of hosting already, and they’re looking at dockerising and containing their own hosting platform for Websites and for services.

You’ve got AWS, which has been in this space since you could probably say day one, because a lot of their compute and resources is based on containerisation and spot type resourcing. So AWS doesn’t have a native operating system and relies on Linux heavily in open source. So they’ve been in this space for a number of years.

But recently we’ve seen Microsoft make a big play into this market. A couple of the features and the functionalities that they’ve added, especially along the API management aspect have made a lot of positive progress. These allow businesses to actually take advantage of modernising applications by offering the usefulness of the application through APIs, and this has made Microsoft more of a market leader than the others. They’ve taken that step forward to drop the cost of utilising Serverless environments, but also provided a gateway for any type of business, whether it’s a small law firm to a big corporate insurance firm to take advantage of them. Whereas the AWS and the Google market still plays very much strategically with the bigger players, because there’s a lot that needs to be configured and put in place, whereas the Microsoft market is very much open to small companies as well as large ones.