Is Fargate for EKS "Serverless"​?

Last week at re:Invent, AWS has announced EKS running on Fargate. Fargate is often described as Serverless offering for containers. If you have used AWS Lambda or equivalent FaaS services like Azure Functions, you might wonder if this is really a Serverless offering? Before answering this question, let's see what Fargate brings on to the table.

What is Fargate?

When Fargate was introduced in 2017, it had brought in a new paradigm of how containers can be run on Cloud. Instead of manually setting up container clusters (using ECS) on top of EC2 instances, Fargate had automated the provisioning of containers as required and eliminated the need for planning and provisioning the compute. Fargate uses lightweight virtualization technology called 'Firecracker' to run the containers within virtual machines, offering strong isolation, security and multi-tenancy, which AWS describes as "MicroVMs". Firecracker is what makes Fargate so unique.
Fargate for EKS

Fargate so far only supported ECS, but had no support for EKS (managed Kubernetes service from AWS). With this announcement EKS can run on Fargate based compute for hosting containers. With EKS on Fargate, setting up an EKS cluster is a matter a single command plus creation of Fargate profile and some IAM tweaking. Once done, EKS cluster can scale based on application scaling requirements as determined by the pod scaling behavior. Most importantly, the scaling is elastic and customers are charged based on the containers provisioned.

Here is an animation from @ArchanaSrikanta that shows how extensible Kubernetes architecture is, and how simple it is to use Fargate.

Is it Serverless?

Serverless computing is a word that is overloaded to each vendors' conveniences. But if we have to scope in words, here is how I characterize 'Serverless'.
  • Cost - Strictly usage based pricing
  • Dev Experience - Source-centric applications
  • Scale - Truly elastic scaling
  • Setup - No infrastructure provisioning

More the boxes checked by an offering, better Serverless it is. How does Fargate or Fargate for EKS fare against these characteristics?

  • Fargate pricing is based on containers provisioned by it to handle the requests, and definitely not based on the actual requests. This box is not checked.
  • Containers are very close to source, so I would say this box is checked.
  • Fargate scales containers elastically up and down as needed. This box is checked.
  • Creating an EKS cluster on Fargate is fairly simple. There is no need to plan and setup the underlying compute, network and storage. This box is checked too.

Fargate for EKS is quite close to be called as Serverless offering, except for the way it is costed.
Fargate for EKS is quite innovative in how Kubernetes is extended to run on a completely different compute infrastructure, albeit with a few limitations.

AWS claims they run 80% of all containers that run on public cloud. This claim show the dominance of AWS. With such dominance, Fargate for EKS is all set to be a leading Container Serverless offering for Kubernetes. Are there any other Container Serverless offerings for Kubernetes that compare to Fargate for EKS? That would probably be the next topic to write about.

This post originally appeared at
Most enterprises are struggling to innovate and are not investing enough into new technologies to counter disruptive measures. As a result, in a few years, their products are becoming irrelevant in the market. Examples of Blackberry, Nokia, Kodak, Blockbuster have been quoted numerous times. Technology has lowered the entry barrier to innovation, and new entrants can easily disrupt the entire business models in a matter of months. Geographical borders do not guarantee any competitive advantages anymore. Any breakthrough or new idea travels around the world in a matter of minutes.

For an enterprise embarking on the digital transformation, following key things need to fall in place for building digital capability within the enterprise:

  • Business vision – digital transformation initiatives are cross division or cross-functional, as a result, having a common business vision to drive the transformation is very important
  • Leader led – the digital initiatives need to be top driven and led by a senior leader. The cross-functional team should report to the leader
  • People skills – cross-functional team needs to be staffed with people that bring varied experience and are adept at defining and refining digital journey focused on the customer experience. In case, the team is relatively new to digital, then hiring for appropriate skills need to be undertaken. Existing team members can undergo training for required skills. People that bring multiple skills to the table are core to the team
  • Platform approach and Automation - when looking for technology choices in implementing the identified projects, start thinking in terms of platforms instead of systems. Digital transformation is not about one project or program, the continuous iterations and interconnection of functionality, requires platforms that provide the requisite services to build such experience in a rapid manner. The implementation team should not get bogged with scalability and performance issues, things that mar the customer experience. The platform should handle the scalability and performance aspects of the application. Similarly, horizontal capabilities like data analytics, data aggregation, content services, logging services should be viewed and built as platform services. Again, if platform building skills are not available in-house, they need to be hired from market to jumpstart the process. Another key capability is to automate the Development to Operations application lifecycle (DevOps). When you working on scale, manual processes and intervention is error prone and not at all feasible. Investment in automation of all the DevOps process is an important part of the building digital capabilities
  • Open source software – in the digital products and services space, given that the product(s) can become viral overnight, scalability and performance of the product/service is a key criterion. Use of commercial proprietary products limits the ability of the enterprise to scale the product. Licensing cost is also the key factor for handling user base in excess of 100 thousand. Open source software allows benefiting from the larger community, cutting down on the costly license, use of commodity hardware and ability to dig deeper into code (if required). The platform approach is executed using a combination of open source software.

Enterprises threatened with changing business models have already built or started building digital capabilities. Companies like Walmart are no longer just brick and mortar players but giving tough competition in the online e-commerce space also. Investment into future is required for enterprise to reinvent them for the digital era.

This post originally appeared at
For enterprises disrupting the marketplace, technology is the key enabler that is helping them create new business models and processes. Open source software has played an enabler role in developing these new scalable and high-performing systems. From the digital transformation perspective, following four levers play an important role.

  • Connected – with the advent of IoT everything is getting connected. All this data is being transmitted over mobile and fixed bandwidth networks where the enterprises are making sense out of it leading to real-time decision making. New service models are being created around data. E.g. Rolls Royce aircraft engines capture data around the engine performance which can be given as subscription to airlines for predictive maintenance. This data can also be used to develop models for better engines.
  • Data – All the data from Things, device location, social interactions etc needs to be captured, processed and analyzed for patterns, behavior discovery to allow better predictions and decisions to be made.
  • Automation – Given the scale of the systems and data being emitted/generated for analysis, automation is the key enabler. Traditional technologies need to be combined with Artificial intelligence and machine learning models to increasingly create systems that can work autonomously. Lots of supervisor roles can be automated using the AI models. E.g. Uber has eliminated the complete middle management role found in a traditional taxi services company.
  • Users First – All the data and automation need to be used to build and provide contextualized products and services for the users. Enterprises or intermediaries can get direct access to users/customers via mobile apps. Users can have fully transparent access to their data and enterprises can offer services around that. E.g. Tesla have built a model for direct selling of cars to the end customer bypassing the franchise/dealer network. This allows them to understand the customer behavior and usage of car. They can sell additional services based on the user context.

It is still early days of digital transformation. Lots of enterprises are taking steps with digital as digital marketing which is a very narrow and siloed focus. Digital transformation is about rethinking the business, operational and technology processes (e.g. customer onboarding, customer cross/up sell, customer service, core products/services etc). Technology is the key enabler and drives the entire business model in the digital economy. Every company is being transformed in a technology company where technology is enabling all these digital interactions.

This post originally appeared at