In working on making their service work for me, the folks at VONAGE recommended PingPlotter -- and I must say it's terrific.



I am supposed to have 512kb/sec symmetric DSL/C (on its own dry copper pair/no phone on the line); Paradyne MVL modem... Watchguard SOHO fiewall...



I've been trying to solve a really awful problem with the Vonage VOIP service.... If I do ANYTHING that downloads while i'm on the phone, the incoming voice gets chopped beyond usability.



I've been PP-ing a server in my ISP's net -- and I get moments when the RTT's are in the hundreds of MS... I can cause the RTT's to go to heck if I download a big file; if I upload I see a degradation but nothing like the impact of a download.



My question: does anyone have a similar PP image? (OR: if you look at JUST the RTT to the FIRST router at the ISP end of the DSL line, what are the Min/Max/Average times?



Thanks in advance



Dennis


<pre>
Target Name: www.mv.com
IP: 199.125.85.33
Date/Time: 05/02/04 6:39:47 PM to 05/02/04 8:59:34 PM

Hop Sent Err PL% Min Max Avg Host Name / [IP]
1 499 0 0.0 0 9 1 [10.15.2.1]
2 499 0 0.0 16 2185 46 gw-dslc-1.mv.net [207.22.56.1]
3 499 0 0.0 16 2170 37 copper.mv.net [199.125.85.33]</pre>

Hi, Dennis.

Last weekend, I whipped up the following tutorial article (but I've not added it to the index yet) that addresses a substantial amount of what you're asking about here.

Have a look here:

http://www.pingplotter.com/tutorial/ScenarioSaturatedPipe.html

After you've read through that, feel free to come back here and post any additional questions / observations that you might have.

- Pete

Pete,

Many thanks for the fast reply... the post does help a lot -- but I'm not sure the behavior I'm seeing "looks" like that. If I UPLOAD a 9Mb file to an FTP server at my ISP it takes about 3 minutes - and the RTT's don't go all that wacky; but if I DOWNLOAD the same file (which takes the SAME 3 minutes!), I get a wildly different result.

Here's the DOWNLOAD info:
Target Name: www.mv.com
IP: 199.125.85.33
Date/Time: 05/02/04 9:20:07 PM to 05/02/04 9:24:12 PM

Hop Sent Err PL% Min Max Avg Host Name / [IP]
1 50 0 0.0 0 3 1 [10.15.2.1]
2 50 0 0.0 2 7 3 [192.168.102.1]
3 50 1 2.0 18 1243 768 gw-dslc-1.mv.net [207.22.56.1]
4 50 1 2.0 18 1212 765 copper.mv.net [199.125.85.33]

Here's the summary on the UPLOAD of the same file:

Target Name: www.mv.com
IP: 199.125.85.33
Date/Time: 05/02/04 9:24:27 PM to 05/02/04 9:28:02 PM

Hop Sent Err PL% Min Max Avg Host Name / [IP]
1 44 0 0.0 0 5 1 [10.15.2.1]
2 44 0 0.0 2 8 3 [192.168.102.1]
3 44 1 2.3 18 424 148 gw-dslc-1.mv.net [207.22.56.1]
4 44 0 0.0 18 255 141 copper.mv.net [199.125.85.33]

I can absolutely understand saturation - but I'd expect that it ould EITHER be truly symmetric (which my DSL service is spposed to be) OR have DOWLOAD be more durable since most consumer network services are tuned up to be symmetric...

Have you seen "one of these"?

Thanks again,

Dennis

Hi, Dennis.

To be honest, I've not spent any time eyeballing any symmetrical DSL latencies, so it's hard to know for certain whether they should look the same or not. The nuances can start to get pretty complex when you start to compare things like this - and you might get focused on things that don't really matter.

The bottom line is that when downloading files, your latencies go to crap. This *could* be a prioritization issue within your DSL provider.

You'll notice on the link I gave you a bit ago, that the maximum latencies hit a bit over 800, with an average of 484 - 600 (depending on hops). That's a bit lower than what you're seeing, but not significantly (except for the Maximum, which honestly isn't very important unless you're seeing the maximum number regularly - which you're probably not or you'd get averages closer to that).

It would probably be more helpful to actually post some pictures with time-graphs, rather than just min/avg/max values - I have a hard time speculating without seeing a bit more of the "picture" than the text summary shows.

Basically, though, what I see from your performance isn't a whole lot different than what I was getting. The difference between 512K connection and a 640K connection might account for some of it, and variations in configurations might account for the rest.

What you're really looking for is some guarantee of Quality of Service (QOS) for a specific type of network packet. What you *want* is voice traffic to be prioritized higher than download traffic. This requires some additional configuration and often additional network hardware to manage. Ideally, your network service provider has some QOS equipment / software in place to allow them to prioritize your VoIP traffic differently than your other traffic. This is what *Serious* users of VoIP do - they set up agreements with their network providers to guarantee delivery of voice traffic.

I'm not aware of DSL providers that do the same thing - but it's certainly possible. I'm not an expert on the current crop of DSL hardware - maybe there's something already in place for doing this.

You might find out from Vonage what maximum latency their applications can have while still maintaining the quality of voice service you expect. If this is 250ms, then you're not going to be able to get that while download a file without some intervention from your DSL provider. If it's 500ms, you can possibly do some "tweaking" of your network to make it work (packet sizes can have a significant impact on latency, and there are almost certainly some things you can do to decrease your latency).

You will probably want to visit with Vonage a bit to see what they have to say. I can't imagine a situation where you're going to be able to download at full throttle and still maintain a reasonable quality voice conversation without doing some kind of packet scheduling at the bottleneck (your DSL line).

The upload latency is a bit puzzling. Saturating either side of your bandwidth (either upstream or downstream) will cause significant latency problems - and I, too, expect some similarities. Is there any possibility that the upstream packets are significantly smaller than the downstream packets? If you use HTTP to upload / download a file, do you get similar inconsistencies between upload and download speeds?

Unfortunately, I'm *NOT* an expert in minimizing the latency of a saturated connection. Your DSL provider and Vonage will probably be more help than I am on that topic.

You can attach an image to a reply if you'd like to give me something a bit more detailed to look at. If you set your time graph to 5 minutes and then focus in on a period of upload/download, then we can look at the specific latencies that you're seeing here.

- Pete

Really appreciate all the assistance... and I did take all the info/images/etc. and ship them off to my ISP.

I'll attach the image of the simple test (FTP upload followed by an FTP download):

Best,

Dennis

That's a *severe* disparity. I don't think this is normal.

Have you done much modifying of your TCP/IP stack? ie: MTU size, etc? Do you have any other PCs on your network where you can verify the results with a different configuration?

I think your next step is to ask your ISP (ie: mv.com) what they think you should do about this, and possible get Vonage involved. Beyond that, I can't offer much advice to help here.

I'd love to hear continuing followup on this - ie: the response from your ISP, and any change of status that happens.

- Pete