6.1 and 6.2 on qnet ?!?

hi,
Is there any problem to run 6.1 and 6.2 nodes running qnet and connected
together, because since I have two 6.2 nodes running on a net with five
6.1 nodes, the 6.2 nodes can’t talk together correctly through qnet.

Alain.

Alain,

Don’t know if this is related to what you are seeing or not, but I was
trying to copy a 2GByte file (boot partition imag file) from one machine to
another during an upgrade, and was getting estimated copy times of around
1.5 hours. I am using qnx 6.2.1 PE, and both machines are P4s with the new
intel gigabit ethernet chips in them. If I am right about my math, it should
only take a minuite or two to transfer that amount of data at 10Mbits / sec.
(((2000000000 * 10 /10000000)/60) is about 33 min) Since I have the two
machines connected directly with a cross over cable, I would expect them to
be at least 10X faster than that at 100Mbits /sec. It should only take 3 min
at that rate, or 3 seconds at 1Gbit / sec.

I checked my cpu usage, and network utilization indicators on the shelf, and
I was under 2% of the network utilization, and the processor was not pegged
on either machine. It sounds like there is some built in delay in the main
loop of the qnet code that is slowing down transfers, or some other conflict
using qnet in version 6.2.1.

Have you compared transfer rates on your system between two 6.2.1 machines
vs 2 6.1.0 machines??

Thanks,

John

If I am right in my math,
“Alain Bonnefoy” <alain.bonnefoy@icbt.com> wrote in message
news:3EE0A976.6000002@icbt.com

hi,
Is there any problem to run 6.1 and 6.2 nodes running qnet and connected
together, because since I have two 6.2 nodes running on a net with five
6.1 nodes, the 6.2 nodes can’t talk together correctly through qnet.

Alain.

John Eddy a écrit:

Alain,

Don’t know if this is related to what you are seeing or not, but I was
trying to copy a 2GByte file (boot partition imag file) from one machine to
another during an upgrade, and was getting estimated copy times of around
1.5 hours. I am using qnx 6.2.1 PE, and both machines are P4s with the new
intel gigabit ethernet chips in them. If I am right about my math, it should
only take a minuite or two to transfer that amount of data at 10Mbits / sec.
(((2000000000 * 10 /10000000)/60) is about 33 min) Since I have the two
machines connected directly with a cross over cable, I would expect them to
be at least 10X faster than that at 100Mbits /sec. It should only take 3 min
at that rate, or 3 seconds at 1Gbit / sec.

I checked my cpu usage, and network utilization indicators on the shelf, and
I was under 2% of the network utilization, and the processor was not pegged
on either machine. It sounds like there is some built in delay in the main
loop of the qnet code that is slowing down transfers, or some other conflict
using qnet in version 6.2.1.

Have you compared transfer rates on your system between two 6.2.1 machines
vs 2 6.1.0 machines??


It’s said that it should be faster in 6.2.1!?! but I didn’t make any

test. I will see soon.
In fact, my problem was an IRQ conflict.

regards,
Alain

Thanks,

John

If I am right in my math,
“Alain Bonnefoy” <> alain.bonnefoy@icbt.com> > wrote in message
news:> 3EE0A976.6000002@icbt.com> …


hi,
Is there any problem to run 6.1 and 6.2 nodes running qnet and connected
together, because since I have two 6.2 nodes running on a net with five
6.1 nodes, the 6.2 nodes can’t talk together correctly through qnet.

Alain.



\