When I run PingPlotter while downloading a file the lost packet rate goes from zero to a very high rate. Is this normal or the sign of a problem.

I've been using your utility for a few years now and has helped me convince ISPs of problems. Great utility,

Thanks.....

That's relatively normal. Usually, when you download something you should get a big jump in latency (as your bandwidth is being saturated). Sometimes, this will be accompanied by a big jump in packet loss as well.

These symptoms are more true with lower bandwidth connections. As you increase bandwidth, you should see less latency jump and less packets lost, although it will still happen as you're saturating your bandwidth.

Well I'm supposed to have a fast connection, and when it's working properly it's very fast. I have recorded download speeds of >2.5Mb sustained.

The ISP had lightning problems about a month ago, had to replace equipment, and since the problems started the downloads speeds have dropped and are intermittently BELOW dialup. At best all I get now is around 1Mb.

The system we have is the enitire community is wired as a 10MB LAN that carrries data and analog TV. Our individual access is through an interface provided by the ISP that supposedly just keeps the TV stuff out of out PCs. It's not a modem.

Where I'm seeing the intermittent increase in latency is between the ISP's last machine and the first Sprint machine. There are two machines between me and Sprint. The latency between me and the ISP's two machines never increases above 3 or 4ms. I'm starting a 48 hour plot at a 2.5s interval and will post the results for you to see.

The periods of high latency have no pattern to them, they can and do happen at any time of day. May last a few minutes or a few hours.

BTW. Reviewing posts here on your forum where plots have been posted I cannot see the picture. In IE I get a picture place holder symbol and in Mozilla I see nothing at all. Any suggestions.

Thanks you very much.....

Forget to mention. There is no noticeable change in latency while downloading, just the increase in packet loss.

Another question. Would it be better for me to plot just to the troublesome Sprint connection or to some point further away?

You should *always* trace to a real server (it can be a mail server, news server, or DNS server very close to you, though). Ideally, the server you trace to is one that you regularly use and have problems with. This does a few things for you:

1) It gives credability to your data when you use it to complain about problems. If you trace to a router and it has performance problems ... who cares? No one uses a *router* - they go to another destination.

2) It could be that that router wasn't even *supposed* to be doing anything durint the period you traced because the routing tables were selecting a different router to be used. If that's the case, then all your data collection was invalid.

3) Routers can respond completely differently when they are traced to directly, rather than through.

It's almost certain that you can find a server / service very close to you that is normally something that people use (ie: mail, dns, news). Instead of looking for a router, look for one of these.

Pictures are *usually* posted on the users web site, which tend to change / go down after a period of time. Some people send me data - which I then post on our web servers, but most pictures from users posts don't persist over a period of a year or more because of this.

Your plan is good - trace for several days and look for a pattern. I look forward to seeing this.

If you send me a picture in email (or give me permission to mirror your pictures), I can put the images on our servers.

Understand the pix situation.

I'm at another location on the same network now. The network was unavailable for about 40 minutes and just came back up about 20 minutes ago.

So far at least, while plotting and downloading at the same time there no packet losses that trigger the graph. So, maybe they have found something.

Can you tell what difference in download speed can be attributed to PC speed and OS? This system I'm looking at now is an old 300Mhz job running Win95. Right now I'm seeing sustained downloads of about 1Mb on this system.

My system at home is a 2.4GHz unit running WinXP Pro. I'll check the download speeds when I get home.

Thanks for all your timely help.

Great product, Great support....

Another question.

I was actively plotting and tried to insert a comment. I right clicked on the time line where I wanted, added my note, clicked OK and the graph was frozen.

I clicked on reset focus, the graph started moving again and the previous 45 minutes of data became corrupted. My nice 40ms trace now has peaks up to 950ms. Any idea what I did wrong?

Thanks again....

Hmmm. That's one I don't know about.

If you happened to drag (left-click-drag) the graph a tiny bit right at the point you were doing your comment, then the time-graph will stop moving. This doesn't change the way data is *collected* at all, though.

The fact the you could re-focus on the current time makes me think that's probalby what happened.

Now, the corrupted data is very odd. If memory was physically corrupted somehow (either by a program bug or hardware problem), I'd expect to see peaks a lot higher than 950, as the scale on this graph goes to 32767. Having random data in just the lower 2% of this range would be really odd.

Do *all* the hops show this same issue?

Did you save the data so you can send it to me?

Thanks,
Pete

I may have done the left click drag thing. I don't have the data anymore so I can't go back and look at it.

It looks like the outage I mentioned this morning was for some work on the network and the ping times have been in the 40~50ms range to the first Sprint machine. And yes, that was the only place where the latency was flakely.

Now I'm still agonizing over the issue of increased packet loss while im downloading and plotting at the same time.

I checked another machine on the community network and there is no increase in packet loss. The difference between that one and mine is our first hops have different machines. The second hop which goes directly to Sprint is via the same machine.

When this incresed packet losses start there in NO noticeable increase in latency. Being anal retentive I'm just concerned that with the increase in packer loss my download speeds are suffering. What do you think.

And my other question was does a fast, up to date (2.4GHz) WinXP PC have much if any advantage over a much slower (300MHz) Win95 unit.

Thanks for your continued input....

Where are you seeing the packet loss? Can you post a picture (or send it to me via email)?

I think, in general, that a 2.4 ghz machine will probably have at most a very small advantage in download speed if that's the only difference. It's likely that there are other variables, though - though; Similar RAM? Same operating system? Same network settings? Same Network card? Same network cable? Same router on the other end of the network cable? Same network subnet?

There are a huge number of variables here that I think have the opportunity to affect the download speed far more than processor speed. Note that my opinion here is antecdotal, though, and I haven't personally run tests to say what affects processor speed has on download speed.

I need more data before I feel comfortable speculating on issues at play here.

- Pete