To test the performance of the VoIP side of the GP2, I turned to www.testyourvoip.com. This is a free service run by Brix Networks in hopes that it will whet your appetite for their pay-for VoIP performance management products. TestYourVoip.com lets you run two types of tests. The first uses a Java applet to make a VoIP call to Boston, Helsinki, London, Montreal, San Jose or Sydney. This test essentially tests the fitness of your computer and Internet connection for handling VoIP traffic. The second (which requires free registration) lets you make a voice call to a "Golden Phone", which measures the actual voice quality of a call made through a VoIP service.
Note: The website's Java-based testing system apparently only works with IE and not Firefox...but that's another subject.
Using the GP2's default settings, my test using the Java applet scored an MOS (Mean Opinion Score) of 4.2 out of 5, or between "Like calling next door" and "As good as a decent cell phone call", as they put it.
I then made a call to the site's "Golden Phone" and got a score of 4.4. Here's what the site says about the testing results:
"While the theoretical MOS scale tops out at 5.0, practically speaking, you won't get a 5.0 score no matter how good your network connection is. That's because VoIP codecs introduce some amount of quality loss. For example, the maximum MOS score you can achieve with the quality-preserving G.711 codec is 4.4. For the low-bandwidth G.729 codec, the maximum is only 4.2."
I next checked to see if the GP2's QoS (Quality of Service) feature really works. If you have even a moderately busy Internet connection, or an ISP that cranks down your upstream bandwidth to 128 kbps or so, you'll probably quickly discover how bad a VoIP call can be. The GP2's QoS features let you assign a High or Low traffic (data) priority by application (logical port), physical LAN port or the built-in VoIP adapter. (The physical port-based QoS also allows you to enable / disable Flow Control and set an ingress (download) rate limit of 8M, 4M, 2M, 1M, 512K, 256K, or 128 kbps in addition to its High / Low prioritization of outbound traffic.)
Note: The GP2's QoS prioritization features only affect how the GP2 uses the Internet connection between you and your ISP. Think of a nightclub bouncer who always lets the High-priority "beautiful people" in first and you'll get the idea of how QoS priority works. The main difference is that the nightclub is your Internet connection (and there's not really a little burly bouncer inside the GP2...). Once your data hits your ISP's network, all (priority) bets are off.
That's not to say that simple High / Low priority QoS isn't useful. Since the pipe between you and your ISP is relatively small - especially for upstream traffic - you can run out of capacity pretty quickly.
Figure 9 shows that by default, the QoS application and port preference features are off and the VoIP system is given priority when it comes to network bandwidth.
Figure 9: Quality of Service defaults
For my test, I enabled high priority for port 6346, set the Voice section of QoS to disabled, then fired up my gnutella client and started downloading everything in sight - maxing out my Internet connection. This effectively killed the phone system completely, so much so that I couldn't even run the testyourvoip.com tests. Even with no downloads in progress, MOS scores went down to 3.0 out of 5, between "As bad as a crummy cell phone call" and "Like tin cans and string".
Switching back to the default QoS settings, the phone system was back to normal and scoring an MOS of 4.4 again - even after starting more gnutella downloads and adding an FTP download to the mix. What this all means is that by using the QoS features, possibly in combination with the Vonage "Bandwidth saver" feature, you can tweak the system to use the bandwidth you have the way you want. (Note also that the physical port-based ingress rate controls can help keep your kids (or employees) from sucking up all available bandwidth for their downloading sessions!)