Glenn Engstrand

Writing Reactive Microservices for the JVM

I’ve known about Vert.x and Play for quite some time but just never got around to covering them here. I recently decided that it was past time to correct for that error in judgement. For those wishing to build reactive microservices for the JVM, these two frameworks definitely merit your consideration. While you don’t absolutely require a reactive framework in order to develop reactive microservices, it sure does help to adopt one. If you feel comfortable developing microservices that don’t depend on the Spring framework, then read on.

These blogs have been getting more ambitious in scope lately. For the longest time, I would develop the next feature identical rudimentary news feed microservice and blog about how it compares to a previously coded and covered implementation. That type of story usually takes about 6 pages to tell. The last 2 blogs had a wider scope, comparing golang to all previous implementations and comparing both REST vs GraphQL with javascript vs typescript. They both were 10 pages in length each. This blog compares reactive to non-reactive and 2 new frameworks with both each other and with a servlet based predecessor. That is why this latest blog is 14 pages in length.

References

link description
Writing Reactive Microservices for the JVM article below
Vert.x reactive web framework used by feed 11
Play reactive web framework used by feed 12
Google Trends Google trends comparison of Play with Vert.x
Netty server backend used by both feed 11 and feed 12
reactive programming concepts behind reactive programming
feed 6 news feed implementation in Scalatra
feed 11 news feed implementation in Vert.x
feed 12 news feed implementation in Play
Slick library used by feed 11 and 12 for accessing MySql
Scalatra servlet web framework used by feed 6
Circe Library used by feed 11 and 12 for working with JSON
Guice Library used by feed 12 for dependency injection
Gatling Library used by feed 12 for local profile based load testing
Akka HTTP Play can be configured to use this instead of Netty
Java NIO Netty uses this part of the JRE for non-blocking IO

By the Numbers

Here is a summary of the article below in numbers.

Static Code Analysis

per file Lines of Code cyclomatic external
feed mean median std dev complexity dependencies
6 40.15 36 29.47 1,981 110
11 42.33 33 33.58 1,300 136
12 55.48 51 32.92 1,764 177

Performance Under Load

throughput latency percentiles
RPM milliseconds
feed entity mean mean median 95th 99th
6 participant 3,885 6 6 9 12
11 participant 3,035 8 7 14 19
12 participant 1,824 7 7 10 13
6 outbound 9,580 9 9 14 19
11 outbound 14,136 5.3 5 11 15
12 outbound 20,151 4.4 4 8 11

Writing Reactive Microservices for the JVM

Having trouble viewing the embedded document? Feel free to download the PDF for offline viewing.

Post navigation

Comments are closed.