[cap-talk] Dan Bernstein's qmail security lessons paper

[Sandro Magi]

David Wagner wrote:
>> Java has dreadful performance problems:
> 
> Well, "dreadful" is rhetoric that appeals to emotion rather than reason.
> I think it's a good idea to quantify these claims.  How much slower is a
> Java-based web server, using metrics that are an appropriate measure of
> end-to-end web server performance?
> 
> It's my impression that Java is arguably the dominant platform for web
> services, in the enterprise world.  How do you reconcile that with your
> view of Java?

The only real performance bottlenecks for garbage collected languages
are tight heaps as found in small devices, and large heaps with virtual
memory where tracing causes thrashing [1]. Unfortunately, server-side
systems can fall into the latter category. Note that this problem is
endemic to ALL tracing garbage collection algorithms, though heuristics
like generations can reduce the thrashing [2].

Additional hardware can make up any performance deficiency, and the
advantages of using Java or ASP.NET with its attendant tooling and lower
risk probably makes up for any additional hardware costs.

Sandro

[1] http://lambda-the-ultimate.org/node/2552
[2] I think a good region-based memory management solution is
"imminent"; now there's a good typed low-level framework [3] which
addresses all the deficiencies of past systems, so only a decent
inference algorithm is missing.
[3] http://lambda-the-ultimate.org/node/2551