I am trying to diagnose brief slowdowns on a server, and in the process, discovered PingPlotter.

If I understand correctly, these graphs shows that the server (13) is not the problem. I did one with 10 samples, the second with 25, and the third with 106. It looks to me like the only consistent culprit is in the data center, (11 and 12).








Am I readig this right? Is there more to be gleaned?
Thaks much,

Bill

Hi, Bill.

The final destination doesn't look bad, really. The intermediate hops show some latency that doesn't show up at the final destination, which basically means they don't prioritize creating an ICMP reply as high as passing data through (not entirely unusual).

Because the final destination doesn't show much of an issue, it's hard to speculate on what's going on. Without a problem to troubleshoot, we're not going to have much luck finding a problem in the earlier hops either.

We discuss this concept a bit here:

http://www.nessoft.com/kb/24
and
http://www.nessoft.com/kb/2

It looks like you're using 1 minute trace intervals here. That's pretty long. With the bandwidth you appear have (based on your latency), I'd suggest using something a lot more often. For initial troubleshooting, I'd suggest using 15 seconds as a maximum (that means an outage of 10 seconds could be missed), but even better something like 1 second or 2.5 seconds. The bandwidth implications are minimal (56 byte packets X 13 hops = 728 bytes. Returning packets are similarly sized), and you'll be able to catch patterns and outage lengths a lot better.

- Pete

Oh - one more thing.

The gray periods in the hops 10 and 11 graphs are because of a route change. This is probably some kind of a load balancing system. You can see that by going to the Route change panel on the left (drag it open if it's closed) and then clicking on a time to show the route that participated at that point. Scroll through a few of these to see the routers included, and if you decide this is normal, you can "mask" this change, and supress the route change logic for this set of routers. To do that, right-click on hop 10 (or 11) and then select "Add route change mask". This will create a mask and also combine the data so that gray period doesn't appear.

- Pete