This is because in Java land startup speed isn't viewed as something important.

Groovy also has the poorest runtime IMHO. JRuby or Clojure are in a better shape.

I'm confused. Why isn't scala also affected by the loading-time issue? It's running on top of the JVM, too...

It's not apples to apples. They are totally different tasks and frameworks.

They are probably having the same issue Rails 2.x has on it. Each request needs to have several libraries loaded. Scala is probably just using servlets fairly directly which entails a minimal spin up time.

I see, so it's not the JVM loading time, so much as library loading time. Interesting. I would have thought that would hit python even harder.

I love app engine and if it had SSL I'd use it for all projects I do ... BUT: It feels odd to run the speed analysis tool in google webmaster tools and discover that my very simple site is slower than 80% of the sites on the internet...

I use python, btw, so it's still fairly fast, but for some reason even the simplest pages (one datastore request by key and a simple template) are still not "fast" compared to the rest of the internet.

Here's a very good explanation of why python performance is faster than java on GAE: http://stackoverflow.com/questions/1085898/choosing-java-vs-...

I hope you don't think this site demonstrated this. The python site was a proxy. The java/scala websites did more than trivial things.

Of course not. But it's helpful to know that information when you're comparing python to java on GAE regardless.

Alright but for your information that post is almost a year old and GAE/J is now much more mature. IMO, a developer should just use what they are more comfortable with. Sadly, it looks like the groovy framework has not been tweaked to work well on GAE.

I was under the impression that Java was faster (even in startup time)...wow. Hope this summary is correct, as I didn't have plans to swap to the java ecosystem.

JVM is the slowest runtime in terms of startup. But it does run way faster once it has run for enough time and has profiled and optimized a good portion of your application.

In most cases where Java is used on the server, it is expected to startup once and continue running for a long time and host multiple applications. In this JVM's startup becomes a non-issue and its optimizations become a huge benefit.

I agree. And it is also not a real benchmark since it obviously tests different types of applications.