Serverless, this other way of designing and running applications in the Cloud, has gained maturity and is becoming more and more apparent. When Amazon Web Services unveiled Lambda in 2014, many saw the service as a simple way to run scripts without worrying about infrastructure. From a small execution engine for SysOps, Serverless has transformed into a natural ecosystem of solutions. Multiple programming languages are available, including the ability to run a custom container, as well as persistence layers and message buses, all the essential building blocks for creating large-scale applications.
An illustration of this change of scale is Amazon Prime’s recent decision to switch its video streaming platform to Serverless. Frédéric Barthelet, formerly head of Theodo ‘s Serverless Tribe, comments on this decision: “A few months ago, the Amazon Prime team announced that migrating the service to AWS Lambda had made it possible to lower their costs by 90%. infrastructure. The use of a prevalent framework called Serverless shows that in May 2023, there were almost 5 million downloads of this framework. Its alternative, the AWS CDK, is also experiencing strong growth, and a disruptive framework like SST is experiencing exponential growth. »
Serverless services are much more sophisticated than they were five years ago, and users have FinOps means to further lower the operating costs of their infrastructures, in particular by opting for ARM versions of these services or more affordable S3 storage modes where possible.
AWS notably offers solutions to respond to a defect inherent in the Serverless approach: the latency induced by “Cold Start,” that is to say, imposes a delay in deploying the function execution container at the time of his very first invocation.
Alternatives To CSP Offers
If AWS, Microsoft with Azure, and Google have been offering Serverless services for years, other players provide alternatives. Thus, Cloudflare is positioned in Zero Trust network services, application-oriented services, and developer services, in which the publisher has been offering Serverless products since 2018.
“On Serverless, we maintain three major differentiators compared to hyperscalers,” argues Stéphane Nouvellon, Solutions Architect at Cloudflare: “On the one hand, there is no notion of region on Cloudflare: we do not launch a service on a region like Europe, for example. It is our job to carry out the replications if the company is a global player. With its 290 data centers, the provider claims a latency time of less than 50 ms, with 99% of the world’s population connected to the Internet. The manager also mentions the robust integration of hosted services.
“It is straightforward to make our services communicate with each other without unnecessary complexity and without introducing latencies. Finally, Stéphane Nouvellon believes he makes a difference in the life cycle of the application: “The company must be able to integrate the application into its software ecosystem, with observability and operations, and integrate the application into dashboards of the SRI team, which takes care of the availability of services.
More Mature Offerings And Methodologies
In addition to a much richer market offering than in the past, the architectures and methodologies have been refined in order to support critical applications. This is the case of the Netherlands Post, which processes around 1.2 million parcels per day on an EDA (Event-driven architecture) type infrastructure. “We literally process billions of events per month on AWS, and we were able to do this by following the rules that we managed to put in place to evolve towards an EDA-type architecture,”
“We start with a few interconnected microservices, optimized locally. Then we add other services, other nodes, we exchange different types of data, we use different types of encoding, we implement other languages: we end up losing any overall structure. It becomes difficult to investigate production incidents, it becomes difficult to modify or remove a service and we end up having difficulty evaluating all the impacts of a change on a microservice.
An approach that appealed to Forrester analysts who placed Cloudflare as a leader in Edge development solutions. Operating and maintaining applications with dozens of microservices running on Serverless infrastructures remains a real challenge for DevOps teams. Solutions adapted to the characteristics of Serverless are emerging. This is particularly the case for “Infrastructure from code” approaches, the idea of which is to generate the Serverless infrastructure simply by reading the source code of the application.
This involves providing self-provisioning capabilities, self-instrumentation, automatic generation of documentation, self-testing and self-auditing, automatic administration and optimization, and automatic updating. Several start-ups have positioned themselves on this “Infrastructure from code” approach: Ampt and Nitric favor an SDK approach; Model and Klotho push code annotation while Encore and Shuttle rely on a mix of the two approaches while Wing and Dark rely on a new specialized language.
Also Read: How Does It Work, And What Are The Advantages Of A Cloud Server?
