Re: Incorrect Throughput Summaries in UDP?


Actually I was thinking the same thing as I was typing the response. 
However I do not know autoconf at all, so I am not sure when I will change 
that. I also want to add a --disable-threads and --disable-ipv6 arguements 
to the configure script but also do not know how to do that. Is there 
anything else that needs to be changed/added to the configure script that 
anyone can think of? If I mess with it, might as well get as much done as 
possible.

Kevin

On Thu, 27 Feb 2003, Hans Blom wrote:

> It would be nice if that "long long" check could be incoporated in the
> autoconfigure. Now it always has to be changed manulally. Or are there
> problems connected with the autoconfigure of the "unsigned long long".
> 
>     Hans
> 
> On Thu, Feb 27, 2003 at 01:56:30PM -0600, Kevin Gibbs wrote:
> > For I assume you mean redhat 7.2 since I know of no versions for "linux" 
> > other than perhaps kernel versions. It should be automatic. You have not 
> > updated or are not using an old headers.h are you? When the configure 
> > script is running does it say you have a int64_t available? if not does it 
> > say that you have a  long long available and then later that its size is 
> > 8? If you do not have a int64_t but a long long of size 8 then you can 
> > change the headers.h to use long long instead of trying to use int64_t. 
> > The change needs to be made at the very bottom of headers.h where it 
> > defines maxsize_t. Just define it to long long.
> > 
> > Kevin
> > 
> > On Thu, 27 Feb 2003, Yee-Ting Li wrote:
> > 
> > > On Thu, 27 Feb 2003, Kevin Gibbs wrote:
> > > 
> > > > What operating system are you using? It would appear that iperf did not 
> > > > find a 64bit integer to use as the total amount of traffic. As such you 
> > > > overflowed your 32bit integer at 4GB. 
> > > 
> > > linux 7.2 - i compiled from source - is there a compile flag i should set?
> > > 
> > > Thanks,
> > > 
> > > Yee.
> > > 
> > > > 
> > > > Kevin
> > > > 
> > > > On Thu, 27 Feb 2003, Yee-Ting Li wrote:
> > > > 
> > > > > Hello,
> > > > > 
> > > > > We're trying to do some iperf tests through an uncongested network using UDP 
> > > > > streams, reporting the rate every 2 seconds for 1 minute. We're using iperf 
> > > > > 1.6.5 with pthreads and we get some strange reporting of the summary result:
> > > > > 
> > > > > 
> > > > > ------------------------------------------------------------
> > > > > Server listening on UDP port 5010
> > > > > Receiving 1472 byte datagrams
> > > > > UDP buffer size: 2097152 Byte (WARNING: requested 1048576 Byte)
> > > > > ------------------------------------------------------------
> > > > > [  3] local xxx.xxx.xxx.xxx port 5010 connected with 
> > > > > xxx.xxx.xxx.xxx port 32769
> > > > > [ ID] Interval       Transfer     Bandwidth       Jitter   Lost/Total 
> > > > > Datagrams
> > > > > [  3]  0.0- 2.0 sec  204248832 Bytes  816993694 bits/sec  0.015 ms 
> > > > > 41676/180432 (23%)
> > > > ...
> > > > > [  3] 58.0-60.0 sec  197239168 Bytes  788955489 bits/sec  0.012 ms 
> > > > > 47441/181435 (26%)
> > > > > [  3]  0.0-60.0 sec  1559126656 Bytes  207849079 bits/sec  0.012 ms 
> > > > > 1466661/5443627 (27%)
> > > > > [  3]  0.0-60.0 sec  2 datagrams received out-of-order
> > > > > 
> > > > > 
> > > > > 
> > > > > As you can see, the summary rate is much smaller than that reported by the 
> > > > > regular intervals; also, the total number of bytes transfered is about a third 
> > > > > less than the accumulated count.
> > > > > 
> > > > > We repeated the test a few times and the report is similar in all tests.
> > > > > 
> > > > > 
> > > > > Any comments/suggestions appreciated!
> > > > > 
> > > > > Yee.
> > > > > 
> > > > > 
> > > > > 
> > > > > 
> > > > 
> > > 
> > > 
> > 
> 



Other Mailing lists | Author Index | Date Index | Subject Index | Thread Index