measuring throughput with netperf

Mark

unread,

Jan 23, 2015, 6:57:11 AM1/23/15

to

Hello,

I'm doing benchmark test of Linux server using netperf. According to
information at
http://wiki.networksecuritytoolkit.org/nstwiki/index.php/LAN_Ethernet_Maximum_Rates,_Generation,_Capturing_%26_Monitoring#Gigabit_Ethernet_Using_UDP
the theoretical max throughput rate for 1514-bytes packets is 117.35 Mbp/s,
however I'm
getting 957 Mbp/s for UDP traffic of 1514 packets. I generate the traffic
with netperf.

I am not sure how this can be possible. Does netperf return udp/tcp or
Ethernet throughput?

--
Mark

---
This email has been checked for viruses by Avast antivirus software.
http://www.avast.com

Stephen

unread,

Jan 23, 2015, 3:38:54 PM1/23/15

to

On Fri, 23 Jan 2015 06:57:22 -0500, "Mark"
<mark_cruz...@hotmail.com> wrote:

>Hello,
>
>I'm doing benchmark test of Linux server using netperf. According to
>information at
>http://wiki.networksecuritytoolkit.org/nstwiki/index.php/LAN_Ethernet_Maximum_Rates,_Generation,_Capturing_%26_Monitoring#Gigabit_Ethernet_Using_UDP
>the theoretical max throughput rate for 1514-bytes packets is 117.35 Mbp/s,

Bytes / sec not bits.........

>however I'm
>getting 957 Mbp/s for UDP traffic of 1514 packets. I generate the traffic
>with netperf.
>
>I am not sure how this can be possible. Does netperf return udp/tcp or
>Ethernet throughput?

Stephen Hope stephe...@xyzworld.com
Replace xyz with ntl to reply

glen herrmannsfeldt

unread,

Jan 23, 2015, 6:22:25 PM1/23/15

to

In comp.dcom.lans.ethernet Stephen <stephe...@xyzworld.com> wrote:
> On Fri, 23 Jan 2015 06:57:22 -0500, "Mark"
> <mark_cruz...@hotmail.com> wrote:

(snip)

>> the theoretical max throughput rate for 1514-bytes packets
>> is 117.35 Mbp/s,

> Bytes / sec not bits.........

>>getting 957 Mbp/s for UDP traffic of 1514 packets. I generate the traffic
>>with netperf.

OK, but 117.35*8 is 938.8, so 957 is still more.

But you have to send a really large amount of data to get good
statistics on this, to avoid problems at the beginning and end.

If you start the clock when the first packet is received, and end
with the last one, you are one packet off.

(note the rates that ftp has reported for years.)

-- glen

Rick Jones

unread,

Jan 25, 2015, 10:59:19 AM1/25/15

to

There can be bits versus Bytes confusion when information sources are
not consistently using 'b' for bits and 'B' for bytes.

rick jones
--
No need to believe in either side, or any side. There is no cause.
There's only yourself. The belief is in your own precision. - Joubert
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...

Rick Jones

unread,

Jan 25, 2015, 11:00:00 AM1/25/15

to

In comp.dcom.lans.ethernet Mark <mark_cruz...@hotmail.com> wrote:
> I'm doing benchmark test of Linux server using netperf. According to
> information at
> http://wiki.networksecuritytoolkit.org/nstwiki/index.php/LAN_Ethernet_Maximum_Rates,_Generation,_Capturing_%26_Monitoring#Gigabit_Ethernet_Using_UDP
> the theoretical max throughput rate for 1514-bytes packets is 117.35
> Mbp/s, however I'm getting 957 Mbp/s for UDP traffic of 1514
> packets. I generate the traffic with netperf.

> I am not sure how this can be possible. Does netperf return udp/tcp
> or Ethernet throughput?

Netperf returns "to the user" throughput. By default it reports
megabits per second (10^6 bits per second).

Following the link, I see them talking about 119,635,891 Bytes/s,
which would be 957,087,128 bits per second or 957.1 megabits/s.

happy benchmarking,

rick jones
--
It is not a question of half full or empty - the glass has a leak.
The real question is "Can it be patched?"