On the client machine, go to a virtual terminal and check for status information on the ethernet. There should be a VT which allows a small shell busybox session. You can use it to test ifconfig or to a ping to the udp-sender machine and see if the network went down for that client.

The numbers you are talking about are not a problem for udpcast. I believe Alain mentioned a site that had around 400 clients at once. The problems may be network or kernel related. If you are using 2.6 now, does 2.4 work OK?

It is not really related, but in my own benchmark testing, I've found the IDE performance of 2.6 kernel sucks. Alan Cox thinks 2.6 ide driver is a mess:

http://kerneltrap.org/node/5062

--Donald Teed

On Wed, 29 Jun 2005, Lasse Riis wrote:

Hi

Well so far so good, now our gbit switch is finally playing nice. So'i attempted to udpcast all 36 clients simultaneously. This didn't work, usually after a few gigs or so udpcast will output Timeout notAnswered=[x] where x is a random client (34 and 22 so far). So now I'm just broadcasting to 12 to see if that'll work, hoping it might work with 24 too. What culd be the cause of this issue? And can I fix it with some parameters.

Also there's definately a speed issue, they're all gbit NICs and it's a gbit switch, but with 36 nodes they run at ~76 Mbps and with 12 it's ~114 Mbps. With a crossed cable between two machines I can get ~360 Mbps. Any ideas on how to speed things up?

Lasse Riis _______________________________________________ Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast