Skip to main content

Java vs. .NET HTTP performance

2 replies [Last post]
mohammed_qaimari
Offline
Joined: 2005-09-22
Points: 0

HI!

I posted this before on the webservices forum, then I found that it would be better if have posted here:

M$ has changed it's HTTP implementation from a Winsock based implementation to a Kernel level driver , this really great for .NET advocates, I mean what can we afford regarding this issue? please read the details at the following link:

http://www.dotnetjunkies.com/Article/99DD7042-532D-4DB4-8625-1CD8BF422D6...

Regards,

Reply viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
linuxhippy
Offline
Joined: 2004-01-07
Points: 0

Well from the security point of few this is quite catastropic. I guess they did the tradeof because they think the kernel-level implementation will stabilize over time and that its worth the performance-win.

However it clearly shows that their infrastructure (kernel, etc.) seems not to scale well for this kind of load for a traditional implementation. The benchmarks posted are ... well ... the best-case for this driver.
Would be really interresting to see un-biased benchmark with linux and solaris for comparison.

The conclusion of my biased post is that I don't think its worth to worry about such desperate tries to be faster than everybody else, in a world where everybody cries for clean and secure implementations.
Furthermore I guess both ASP.NET and Servlers/JSP do have really high overheads compared to the additional work a userspace-http implementation introduces.

lg Clemens

mohammed_qaimari
Offline
Joined: 2005-09-22
Points: 0

I agree that this is a huge drawback from software security point of view.

I intend to make some benchmarks my self to see whether it worth making such a step or not.