Date: Thu, 06 Jun 2002 10:19:15 -0400 From: Bill Moran <wmoran@potentialtech.com> To: "Patrick O'Reilly" <bsd@perimeter.co.za> Cc: FreeBSD Question List <freebsd-questions@freebsd.org> Subject: Re: Voice QOS issue using ipfw/dummynet Message-ID: <3CFF6F63.2090405@potentialtech.com> References: <000501c20d4e$250f11f0$b50d030a@PATRICK>
next in thread | previous in thread | raw e-mail | index | archive | help
I'm no voip expert, so take all I say with a grain of salt. Patrick O'Reilly wrote: > I am trying to run some voice over an IP link using some devices which > convert voice to IP and back again. The IP stream is then passing over > a 64k data line which has a FreeBSD gateway at each end. Have you tested the 64k line to ensure that you're getting true 64k? Sometimes installation mistakes, interference, etc can cause connections not to live up to specs. Check from the FreeBSD machines. Perhaps your interface cards aren't performing up to snuff. > Everything is fine when the first 2 voice channels are open, but once > the 3rd, then 4th channels are open the Quality starts to degrade. The > voice supposedly only requires about 8k per channel, so even with 4 > channels open the line should only be running at about 50% of capacity. I would get some sort of traffic monitor out of /usr/ports/net and monitor the interface to see what it's really running at. My experience has been that advertising departments will stretch minimum requirements. You may be able to get away with 8k/channel, but with degraded performance. It sounds like 16k/channel + is required for optimum performance. But this depends on the results of your research. Is the voip truely only using 8k/channel? Is the 64k line really getting 64k? If yes to both, then you'll need to look for problems in FreeBSD. > I have built the kernels with HZ=1000 just in case the clock granularity > was a problem, but that did not help at all. Anyway, at present the > ipfw rules are passing the voice traffic without even using a dummynet > pipe, so the HZ setting might be inconsequential. Your ipfw ruleset could be a problem. Poorly written ipfw rules can degrade performance. Try turning off ipfw altogether (or add a "allow all from any to any" rule at the beginning) and see if the problem is still there. If it goes away, work on cleaning up your ipfw rules. > Does anyone have any experience with a similar combination of > requirements, and have some advice for me? Not with voip, but with similar problems with SMB traffic over wireless WAN link. One possible solution (in addition to the testing noted above) is vtund from the ports. It can compress the data going across the wire. We saw ~50% bandwidth improvement with it. YMMV > The gateways are running FreeBSD 4.5-RELEASE. They are not doing any > other work, and the data line was not in use for any other data during > testing of the voice issues. I am not sure what other technical info > might be relevant to this issue, so please ask and I'll fill in as > required. Hardware. It could be an issue with the hardware you're using. Not only the PCs, but your switches, hubs and wiring. What does the architecture look like? -- Bill Moran Potential Technology http://www.potentialtech.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3CFF6F63.2090405>