[Udpcast] udpcast scalability?
Gary.Skouson at pnl.gov
Mon Dec 13 22:27:51 CET 2004
We've been able to successfully udpcast to hundreds of clients pretty
consistently. When we've run into problems, it's usually been some problem
with the network. The command from the server end looks something like:
./udpcast/udp-sender --max-wait 60 \
--min-wait 20 --min-clients 70 \
--pipe ./tarScript --interface eth0 \
--max-bitrate 200M --nokbd \
--mcast-addr 18.104.22.168 --full-duplex --nopointopoint \
I've seen over 400 clients connect at the same time
On 12/13/04 11:31 AM, "George Coulouris" <coulouri at ncbi.nlm.nih.gov> wrote:
> We'd like to use udpcast to distribute approximately 20G of data to ~100
> machines on a daily basis. We currently have 72 machines, but usually
> only 40-50 of them successfully transfer data.
> The command lines look like:
> udp-sender --max-bitrate 45m --full-duplex --min-wait 300 --pipe "tar -c
> -f - source_directory"
> udp-receiver --pipe "tar -x -f -"
> We have a cron job which monitors a directory in NFS for a
> flag file, and when the flag is raised, the sender starts up, and after
> a short random timeout, the receivers start up. The receivers write
> messages to syslog, so I can verify that they're starting up before
> min-wait expires. Lengthening min-wait doesn't appear to have an effect.
> We're on switched fast/gig ethernet.
> Are there any reasons why the sender wouldn't be able to "see" more
> Thanks in advance.
> George Coulouris
> National Center for Biotechnology Information (contractor)
> Udpcast mailing list
> Udpcast at udpcast.linux.lu
More information about the Udpcast