Ping problems

Started by Simon_idnet, Sep 07, 2006, 14:03:30

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Simon_idnet

Please can I ask anyone who is experiencing "ping problems" to email support-at-idnet-dot-net so that we can investigate each in detail? Please can you put "ping problem" in the subject line?

We have been performing engineering works on the network during the past week. Apologies if these activities inconvenienced anyone. We have been upgrading interfaces and changing the pathways through which traffic is routed through our network.

With regards the periodical "ping spike" that people have reported seeing on telehouse-gw.idnet.net - this is due to the different methods used by different 'reporting tools'. If you ping the router directly, the interface that the ping arrives on raises a receive interrupt to the CPU which then allocates resources to generate a response. However, pings sent through a router are switched in RAM without generating a CPU interrupt.

Pings destined to the router are allocated a low priority. Every 60 seconds this router checks its BGP routing table and this task takes around 5 seconds. During that time low priority tasks will take longer to complete. Conversely pings sent through a router are treated as normal traffic and thus take precedence over the BGP Scanner.

You can see this for yourself by opening two command line windows and running concurrent ping traces; one to the router directly and one through the router. e.g. if you type "ping -t 212.69.63.40" in one window and "ping -t 212.69.40.254" (this is one of our servers in Telehouse, reached via telehouse-gw.idnet.net) and watch the output for several minutes. The first window will trace responses to pings that are destined directly to the router while the second window will trace responses that are travelling through that router to the server on the other side. The router ping trace shows the "ping spikes" (the BGP Scanner in operation) while the other does not.

Given that the server ping trace is going through the router at the very same time shows that traffic through the router is not affected - only the pings aimed directly at the router are given a lower priority whilst it takes care of its BGP housekeeping stuff.

Graphical ping tools operate by pinging every host in the path concurrently. Pinging our BGP routers is not a valid test of network performance.

The suggestion to try alternative hardware and/or to check that your firmware is up to date stems from several customers having seen improvements with swapping out their router.

We are now confident that our traffic flows within the IDNet network are optimal. I know that there are a few customers whose connection is not performing well. Especially those whose first hop shows a very high ping time. There is clearly something wrong with those connections and so if you could please contact us we will investigate your setup and line characteristics on a case-by-case basis to try to get to the bottom of it for you.

Thanks
Simon

Related links:
http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a00801c4f48.shtml
http://www.cisco.com/warp/public/459/highcpu-bgp.html


AvengerUK

Ive sent you an email about my ping issues about 30minutes ago entitled "RE: Pings".


equk

I've just sent a email with some info attached, just got online and ingame pings are mental aswell as having massive pl.

The ping is stupidly unstable.
e6400 @ 3.2Ghz 38°C 45°C | ATI X1900XT | P5W DH | ss: linux | osx
migration complete - sync 5mb 500k - stable low ping times

equk

well I'm getting various results, depending on which servers I go on. I'm pretty sure it's still a few routing issues as the ping seems stable on others. I posted in the ping thread, so rather than clutter this one here is a link to my post.

ah yer also...

Thanks for the help so far  :angel:
e6400 @ 3.2Ghz 38°C 45°C | ATI X1900XT | P5W DH | ss: linux | osx
migration complete - sync 5mb 500k - stable low ping times