Search All Forums
Dart Home | C++ Professional Toolkit | Custom Development Reply | C++ Professional Toolkit Topics | Forums   
AuthorForum: C++ Professional Toolkit
Topic: benchmarks

From: Vancouver, BC Canada
Posts: 3
Member Since: 05/20/03
posted May 20, 2003 2:26 AM

i'm building a distributed client-server app in VB using the TCP component of PowerTCP 4.1, and i'm trying to determine how many clients a single server could reliably support.

i'm looking for any benchmarks that have been done with the TCP component, either informally or formally. particularly:

1. # of concurrent connections possible
2. resource consumption as a function of # of connections.

of course, i realize some of this depends on the hardware, and how demanding the application is.

also, any pointers to info on the inherent TCP performance characteristics of windows servers would be appreciated.


K M Drake

From: Utica, NY USA
Posts: 3406
Member Since: 07/14/00
posted May 21, 2003 11:10 AM

I'm sorry, I do not have the formal or informal performance statistics you are asking for. As you say, it is really dependent on the system and OS running the software.

My advise would be to test it on machines equivalent to those the application will be distributed to.


From: Vancouver, BC Canada
Posts: 3
Member Since: 05/20/03
posted May 21, 2003 12:09 PM

That's kind of strange that you folks don't even benchmark your own product, or have any performance-related data. Most software companies (including the very small one I work for) provide basic benchmarks on a typical setup. Most people, myself included, aren't looking for an exact figure, but very ballpark figures to see if the product is within particular operating parameters. This is particularly true for server-based software, which is what I plan to use this product for. IMHO, benchmarks are essential to the marketing of a product like this.
Alex Gladshtein

From: Rome, NY USA
Posts: 131
Member Since: 12/27/00
posted May 22, 2003 12:14 PM

Hello Dave,

Thank you for the question. We do have benchmark information on higher level protocols such as FTP and Mail where our layers can impact performance, but on the very low level, there isn't much that happens that would increase of decrease the performance in any significant fashion. Almost all the impact on the low level is based upon the development environment that the produt is used in and the specifications of the machine. We are just providing a more convenient interface to the underlying Winsock API and if you instantiating a component in ASP will impact memory and CPU usage in a much different manner than VB. There are some standard network limitations that one can run across. I believe that 4k is allocated for every socket and Winsock has a cap at something around 5000 simultaneous connections. Basically, on the low level you will hit Microsoft limitations before our product creates any kind of bottleneck. I hope that helps.

Best Regards,

Alex Gladshtein
Dart Communications

From: Vancouver, BC Canada
Posts: 3
Member Since: 05/20/03
posted May 22, 2003 12:20 PM

thanks.. that's all I was looking for, just some basic ballpark data.
Reply | C++ Professional Toolkit Topics | Forums   
This site is powered by PowerTCP WebServer Tool PowerTCP WebServer for ActiveX