Hello,
If udp-receiver is writing to a pipe and the pipe fails, udp-receiver
will incorrectly exit with a returncode of zero. This is easily tested
with a trivial pipe program that copies stdin to stdout, but exits with
a nonzero status upon encountering eof.
Adding the following to startReceiver() in udpr-negotiate.c is
sufficient to fix this behavior:
<---- cut here
int ret;
...
if(pipePid) {
ret=udpc_waitForProcess(pipePid, "Pipe");
}
...
return ret;
<---- cut here
Regards,
George Coulouris
National Center for Biotechnology Information (contractor)
I got a Dell powerconnect 2724 gig switch, and have been testing
it, but have not gotten the results I had hoped for.
Have been trying a number of things, and only about to get
speeds similar to using a 100M switch??
When transfering with 1 sender and 1 receiver I can get speeds of
between 300M and 400M with no retransmits. Set max bitrate to
400M to avoid timeout messages. Have done that with two
different clients with same results.
When sending to the same two clients at the same time (or
sending to 19 clients in another test) I only can get a speed of
about 70M to 78M at best. With no options, it drops to 1M to 2M.
Setting the max bitrate seems to reduce the rexmits, but the
setting of 80M. I've tried it with --fec 8x8/128 and without, and
doesn't seem to make much difference.
It might be the something in the switch, but I have gone in, and
disable the storm protection settings that I had though might be an
issue early in the process?
The switch does have some limitations. Had to use an old 98
machine with old firefox to access its web interface, since it
doesn't seem to work with newer browsers. Might just be this
switch doesn't do well with multi-cast.
+----------------------------------------------------------+
Michael D. Setzer II - Computer Science Instructor
Guam Community College Computer Center
mailto:mikes@kuentos.guam.net
mailto:msetzerii@gmail.com
http://www.guam.net/home/mikes
Guam - Where America's Day Begins
G4L Disk Imaging Project maintainer
http://sourceforge.net/projects/g4l/
+----------------------------------------------------------+
http://setiathome.berkeley.edu (Original)
Number of Seti Units Returned: 19,471
Processing time: 32 years, 290 days, 12 hours, 58 minutes
(Total Hours: 287,489)
BOINC@HOME CREDITS
SETI 11810129.071646 | EINSTEIN 7213450.559851
ROSETTA 4201950.390797 | ABC 10802915.987447
I was just looking out for drivers for NIC and Hard Disk. Their model
numbers are as follows,
NIC:
1) Realtek RTL8168C(P)/8111C(P) Family PCI-E Gigabit Ethernet NIC(NDIS
6:20)
Harddisk:
1) WDC WD2500 AAJS-60M0a0 ATA DEVICE
2) Hitachi HDP725016GLA380 ATA DEVICE
Can you suggest me any other driver which can be compatible with this model
of the hardware.
--
Regards,
Rahul Rao
Msc-CA, SICSR
I just got a 24 port 1GB switch and my first attempt definitely didn't
give the results I was hoping for. Perhaps there are some tuning
options that might improve the options.
First, I created an ntfsclone lzop compressed image of the
Windows partition that is uploaded to one of the machines
connected with the 1GB switch. That took about 11 minutes.
I then made sure that all the machines had the nics running at 1G.
Originally, two machines were only connecting at 100M, but
replacing the connector on 1 and just recrimping the other got
them all to connect at 1G. The cables are all Cat5e, but they are a
few years old.
I have a script that using udpcast to transfer this image file to all
the other 19 machines in the lab connected to the machine. I was
hoping to get a better speed when switching to the 100M to the
1G, but wasn't looking for a 10 times increase, since I knew that
the hard disks only will do about 55MB/ second accordng to
hdparm. But in my first test it took 31 minutes to transfer the file
using udpcast??
Image file is about 12GB.
Here is the end of the listing.
bytes= 12 221 254K re-xmits=32506028 (378.1%) slice=0112 -
15
Timeout
notAnswered=[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18]
notReady=[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18]
nrAns=0 nrRead=0 nrPart=19 avg=252
bytes= 12 229 698K re-xmits=32529469 (378.2%) slice=0112 - 3
Transfer complete.
Didn't notice any machine giving errors, since it seemed to always
list all 19 machines. The re-xmits seems to be an issue, but
perhaps some other setting with the Linux (Fedora 16) or udpcast
options.
The script is run from the machine machine, and it links to the
other 19 machines and then transfers.
Partition /dev/sda6 is is a ext3 partition that is separate on all
machines to store images, that can then be used to reimage
individual machines with no network traffic.
if [ -e /sda6/d7-xp.lzo ] ; then
echo "SDA6 already mounted"
else
sudo mount /dev/sda6 /sda6
fi
cmd="hostname && mount /dev/sda6 /sda6 && cd /sda6 &&
./udp-receiver --no-progress --file d7-xp.lzo && cd / && umount
/sda6 "
#cmd="udp-receiver --file d7-xp.lzo "
plink -t 192.168.7.201 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.202 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.203 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.204 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.205 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.206 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.207 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.208 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.209 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.210 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.211 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.212 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.213 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.214 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.215 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.216 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.217 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.218 -l root -pw xxxxxx "$cmd" &
plink -t 192.168.7.220 -l root -pw xxxxxx "$cmd" &
sleep 20
cd /sda6
time udp-sender --full-duplex --file d7-xp.lzo --min-receivers 19
--min-wait 9
I'm hoping there are some more experience with using udpcast in
1G environments.
Currently running another test with sending to just two machines,
but seems be showing similar results.
+----------------------------------------------------------+
Michael D. Setzer II - Computer Science Instructor
Guam Community College Computer Center
mailto:mikes@kuentos.guam.net
mailto:msetzerii@gmail.com
http://www.guam.net/home/mikes
Guam - Where America's Day Begins
G4L Disk Imaging Project maintainer
http://sourceforge.net/projects/g4l/
+----------------------------------------------------------+
http://setiathome.berkeley.edu (Original)
Number of Seti Units Returned: 19,471
Processing time: 32 years, 290 days, 12 hours, 58 minutes
(Total Hours: 287,489)
BOINC@HOME CREDITS
SETI 11801304.490587 | EINSTEIN 7206065.819851
ROSETTA 4194793.509365 | ABC 10747635.646984
