[erlang-questions] Erlang is the best choice for building commercial application servers

Richard O'Keefe <>
Thu Mar 15 00:27:14 CET 2012


On 14/03/2012, at 11:06 PM, Jan Burse wrote:

> Shahrdad Shadab schrieb:
>> _Java performance is nowhere near Erlang.
> 
> Where is the evidence? I don't know
> whether Joe would approve with this
> statement?

Java is a language with several implementations; even within a single
implementation the -client and -server options will give you different
performance.

I have no shortage of examples where Java is slower than the local
copy of 'mawk'.  (Yep, rewriting AWK into Java and adding all the
type declarations and so on can make it *slower*.)  I also have no
shortage of examples where Java is faster than mawk.  It all depends
on what you are doing and how you are doing it.

Early publications about Mercury showed it doing list manipulation
usefully faster than hand-written C, which was queer (but true!)
because it compiled via C.

I suspect that matching binaries is likely to be faster in Erlang than
in Java, simply because it is not something Java implementors have
ever been focussed on.  I can pretty much guarantee that computing
Singular Value Decompositions in Erlang is going to be *horrible*
compared with Java (although the Java SVD code in a package I bought
turned out to be 8 times slower than C).

There is often surprisingly little connection between the speed of a
_language_ (implementation) and the speed of _systems_ built using it.
In this context, it's worth pointing out that there are some really
good profiling and monitoring tools for *both* Java and Erlang.




More information about the erlang-questions mailing list