ASYNC / Concurreny Communication Distributed Systems
HTTP or asynchronous communication in microservices? This question is frequently repeated and discussed. Obviously, HTTP-based communication is easier for developers and architects. Even if your developers have no prior experience with microservices, they will probably understand how to implement an HTTP service well. While asynchronous communication has a lot of advantages that allow to design and implement really robust microservices system, they also bring new challenges not so obvious for people who didn’t have a chance to work with such an environment before.
In FLYR we mostly have HTTP-based Inter Process Communication (IPC) in our infrastructure. At some point, we realized that to provide the functionality required by our product we needed something more flexible and more… asynchronous. We designed and implemented a Python library to facilitate switching to asynchronous IPC, supporting one- or two-way or even single-request – multi-response communication. An important thing in the design process was to provide developers having HTTP experience a tool that would ease the process of switching to asynchronous communication. Consequently, to switch an HTTP server-side endpoint to asynchronous IPC is a straightforward task.
We selected Kafka for our message broker, not surprisingly by comparing its performance reports with our very rough, but no less demanding performance requirements. But we also took care to hide the details of the broker logic from developers. Yes, we don’t use all Kafka capabilities, if you need e.g. Kafka Streams, you will have to use another solution. But we can decide what capabilities are used in our microservices and how we can make changes in the way our services communicate in a single place. There are also hooks, Kubernetes health checks and more with a lot of flexibility available out of the box.
We plan to opensource our library. At the moment of writing it ‘opensourcing’ is still a work in progress and we didn’t have sufficient time to do it due to strict time constraints we have on delivering functionality to our customers, but we hope to be able to do it soon. In this talk I would like to describe how we solved particularly important problems, what solutions we developed, how we use them and what problems still need to be addressed by developers. In other words, I would like to describe you the journey we made from HTTP to Kafka-based microservices.
Type: Talk (30 mins); Python level: Intermediate; Domain level: Intermediate
Informatics specialist by passion and by profession, a PhD but primarily an engineer interested in distributed systems and software development, with special passion for programming and for Ruby. After 15 years of academic work now developing microservices in Python for FLYR Inc. Member of core team of Rzeszow Ruby User Group (http://rrug.pl)