stutter and other quality problems

Having issues with your magicJack? Post here, and we'll try to help you out!

Moderators: Pilot, Bill Smith

Post Reply
Stewart
Dan Should Pay Me
Posts: 663
Joined: Tue Nov 13, 2007 2:58 pm

stutter and other quality problems

Post by Stewart »

In the past few months, MJ voice quality has seriously degraded. I am writing not to complain, but rather to provide some useful information, so that MJ can find these problems and fix them. I hope that other members will add additional details, to aid the troubleshooting process. IMO at least three distinct faults are contributing to the poor experience:

1. On a small percentage of calls, the incoming audio is so bad that conversation is impossible. However, if you hang up and immediately redial, the new connection is almost always ok. On a test call I made this morning to 415-421-0020 (milliwatt), 40% of packets were lost; see screen shot below.
Image


But when the call was retried (reconnected within three seconds), all packets arrived in a timely fashion.
Image
I don't know what may be causing this, but I've only seen it on the Nashville proxy/media servers. The two calls sourced media from the same IP and AFAICT were routed through the same carrier.

2. At certain times of the day, quality becomes poor for several minutes, then returns to normal (it's usually mediocre at those times). I believe the cause is non-voice traffic overloading your Internet connection at Nashville. Here is a graph showing traffic to (blue) and from (green) XO Internet, covering roughly 8AM PST Friday to 5PM Saturday:
Image
I don't know what the blue spikes are, but they're clearly administrative traffic, not phone calls. Perhaps you are uploading to a remote backup site, or other POP sites are downloading to synchronize their databases. During light traffic times, these transfers go at > 5 megabytes per second! When traffic is heavy, the pipe is already nearly full, and they cause trouble. I made a 55-minute test call at 10:45AM this morning, about 10 minutes before one of these transfers was expected to start. Here is a plot of the number of incoming voice gaps present in each minute of the call:
Image
This problem should be easy to fix; simply rate limit the administrative traffic. If reduced by a factor of ten, though each transfer would take about an hour, there would be no serious impact on voice quality. (You guys need QoS worse than your users ;))

3. There is quite a bit of residual packet loss from Nashville, even when traffic is modest. At 10:45PM this evening, I made a 30-second test call. Although your Internet connection was passing only about 2.5 MB per second, 1.7% of my incoming packets were still lost. To confirm that this was not caused by my Internet connection or your carrier, I repeated the test immediately, using your San Francisco proxy. Not a single packet failed to arrive.

4. A general comment: if you could distribute traffic more evenly among your POP sites, this would greatly reduce or eliminate the above problems. You need to do that soon anyway, if you expect to grow.
Post Reply