What are the cost / performance benefits of using golang in GAE

Question

In terms of quota / usage limits per instance, is there a significant improvement / advantage when using golang in a Google appengine GAE application instead of another suggested language that works in GAE, for example, python, java, php, or do they all behave the same?

Or, in principle, any instance, regardless of the language used, behaves the same and can handle just the same number of maximum requests / sec per instance, given that this relates more to the “GAE load balancer” or to the infrastructure rather than to the programming language used, can the same logic apply to memory, processor usage?

Answer 1

App Engine has no explicit restrictions or restrictions that apply only when using a particular language. However, languages ​​and their technologies may imply certain restrictions, for example, the Java virtual machine instance itself requires significantly more memory and has a significantly longer startup time (even if warm-up requests are enabled) than starting the built-in Go web server, so in the case of an instance Java will have less memory left for Webapp itself to allocate and use (for a specific plan / type and instance).

I have no specific measures to compare, but (in the case of Go):

"The code is deployed in its original form and compiled in the cloud ... Go is the first real compiled language that runs in App Engine. Going to App Engine allows you to deploy efficient, CPU-intensive web applications." ( source )

If you think about it, all languages ​​in App Engine are interpreted (including Java, which is bytecode interpreted by the virtual machine), while Go is compiled and works as platform dependent native code. That should already say something about performance.

For a “case study”, check out the following blog post:

Zero to Go: Launches on the Google homepage in 24 hours

This blog also reports on the performance of a real-world application used by millions:

This chart, taken directly from the App Engine toolbar, shows the average request latency during startup. As you can see, even at boot time, it never exceeds 60 ms, with an average delay of 32 milliseconds. This is viciously fast, given that our request handler processes images and coding on the fly.

App Engine uses a web server, which is included in the standard Go library to serve your application, which also means that you can easily port your Go web application to the App Engine and that you know exactly what to expect from a web server for your application in App Engine

Found Official Python, Java, and Go time comparisons

The state of the App Engine system can be considered an official and good comparison base.

You can click on any cell belonging to a specific day and language and get detailed historical statistics for static and dynamic GET delay (both protected and unprotected), error rate, CPU usage / delay. These statistics are measured on an instance that is ready and ready to serve.

Analyzing it during the day of January 27, 2015 , here are excerpts from Go , Java and Python :

• Dynamic delay is approximately the same for all
• CPU latency (for calculating the 33 ^rd fibonacci number) is best for Java, then Go and the slowest is Python.
• The download time for a static file is about the same, but Go is the fastest.