Windows Server - Technology

The Blog for IT solutions from Microsoft. By Richard Wu

Optimizing Remote Access

One of my friend got problem on VPN performance in his server, so i refer him to look at this articles about optimizing the Remote Access connection:

http://www.microsoft.com/technet/archive/itsolutions/network/maintain/optimiz.mspx?mfr=true

Here is part of it:

Typical Bottlenecks in PPTP Configurations

PPTP connections suffer from two primary types of bottleneck: insufficient Internet bandwidth and CPU load on the PPTP server. The bandwidth problem is easy to conceptualize, since your Internet connection has to be shared between PPTP traffic and everything else that's normally carried over it. You can easily calculate the theoretical maximum number of concurrent PPTP users by dividing your available Internet bandwidth by the average connection speed of your PPTP users. For example, an ISDN BRI line (about 1.5Mbps) could potentially support about 23 64Kbps ISDN users running at full tilt. If you're willing to assume that each PPTP user will only use half as much bandwidth, then in theory, you can support twice as many users.

The additional CPU load added by PPTP is easy to quantify, but hard to predict. If you have a good performance baseline for your servers, you'll already know what your standard server CPU load is, so checking the Processor object's % Processor Time counter will tell you fairly quickly how things stand. You can also watch the Processor object's % Processor Time counter on the RASMAN instance to see how much CPU time RAS itself is using. Overall, though, you'll probably find that the additional overhead is negligible until you get more than 15 to 20 PPTP users on a single server. Since NT will allow you to add as many PPTP ports to your server as you want, be forewarned that allowing hundreds of concurrent ports may impose a speed penalty.