Hi,

I recognized, that using ftp, apache ab or iperf I only get max 60% of 
the theroretical throuput of the gigabit ethernet.

I am using one SF440 (s10s_u2wos_09a) and SF1280 (s10s_u4wos_12b) and 
tested towards zones and/or global zones. In the middle there is only a 
cisco 2960 which shows me the same on the port statistics - only 60%.

Searching the web I saw several network test that sun-sun is one of the 
slowest connect dell-sun ibm-sun all is much better.
But I believe there must be a reason and solution available?

Thanks

Sebastian

Sebastian Fontaine <sfontaine@euroscript.lu> wrote:
> Hi,
> 
> I recognized, that using ftp, apache ab or iperf I only get max 60% of 
> the theroretical throuput of the gigabit ethernet.

I've not seen apache handle gige speeds yet, even with large files. FTP 
can do it if you write the file on the receiving side to /dev/null or a 
memory disk, and have the sending side have it cached in memory already.

In general, no one program is going to saturate a gige connection in the 
first place.

Try the program ttcp. It's able to generate and sink traffic faster than 
bloatware like apache and better for testing connections.

> Try the program ttcp. It's able to generate and sink traffic faster
> than bloatware like apache and better for testing connections.

iperf, mentioned by the OP, is pretty much the same thing.

when diagnosing netperf (yet another choice :) performance, things I
look at are:

*) retransmissions
*) cpu utilization of all the individual CPUs
*) socket buffer/TCP window sizes

rick jones
-- 
web2.0 n, the dot.com reunion tour...
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

In article <6p23dmF5sv5pU1@mid.individual.net>,
	Sebastian Fontaine <sfontaine@euroscript.lu> writes:
> Hi,
> 
> I recognized, that using ftp, apache ab or iperf I only get max 60% of 
> the theroretical throuput of the gigabit ethernet.
> 
> I am using one SF440 (s10s_u2wos_09a) and SF1280 (s10s_u4wos_12b) and 
> tested towards zones and/or global zones. In the middle there is only a 
> cisco 2960 which shows me the same on the port statistics - only 60%.

That would be a pretty typical limit for a singe disk
throughput, which might be your limiting factor.
Not enough detail to comment much further.

-- 
Andrew Gabriel
[email address is not usable -- followup in the newsgroup]