Re: Different throughputs in two directions

Rick Jones (raj@cup.hp.com)
Tue, 23 Jun 1998 14:31:04 -0700

Edwin Law wrote:
>
> Hi,
> I wonder if anyone has an explanation why I sometimes get different
> throughputs of data transfer in the two directions? For example, here
> are some average throughputs I got from the FTP program on the Sun's.
> Both the server and client are using the same socket buffer size and
> application data buffer size. No file cache was present. The results are
> significantly different for the smaller files.
>
> File size (get) (put)
> 5120 0.763618 2.57051
> 51200 4.10513 3.48977
> 512000 7.06257 7.40242
> 5242880 8.44075 8.52563
> 52428800 8.45311 8.51249

There are differences in when the timing starts in FTP get versus put,
wich means that for small transfers, there can be a big difference in
the measured times. Also, different systems can have different socket
buffer sizes, which can affect perforamnce differently depending on
whether the large socket buffer is the one sending or receiving.

> Should Netperf also include some kind of "server's side" timing, like
> TTCP does?

I think the timing should be relative to the system sending the data,
and it should include the time it takes to ensure that all the data has
gotten to the other side (hence the waiting for the two shutdwon calls).

rick jones

-- 
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to email, or post, but please do not do both...
my email address is raj in the cup.hp.com domain...