Ethernet Throughput

[mike_guy29]

Hi all,

I am just wondering what the actual throughput rates are that you get on various Ethernet links (10Mpbs, 100Mpbs, 1Gig)

I am not talking about the amount of data in a packet or frame, just raw bitrate...

Thanks in advance.

Mike

[Giuseppe Larosa]

Hello Mike,

ethernet links have no frame structure so to distinguish between frames a silence is needed.

The minimum silence is called inter-frame-gap.

There is also the 8 bytes preamble that allows a receiver to synchronize.

the overhead of interframe gap and pre-amble cannot be reduced to less then 20.2 bytes.

so as a rule of thumb:

ip size: L3 size

ip size +18 bytes: L2 frame size

L2 frame size+20,2 = equivalent on wire.

Using this and taking mininum L2 frame size is 64 bytes provides number like

10 Mbps -full rate 64 bytes frame:

14845 fps

100 Mbps

148456 fps

1000 Mbps

1484560 fps

this is for one direction only

similar calculations can be done for bigger frame size.

Hope to help

Giuseppe

[kishan1984]

it depends on how much ur server can pump.there is no limit a gig interface can pull data upto 1000 mbits in both direction.

[mike_guy29]

Hi,

Thanks for that. So the interfaces would be pushing data through at the 100Mbps rate but it would be the IFG and the preamble that would reduce actual data throughput. I guess then that frame size would have an affect on the throughput of data as lots of little frames would require more IFGs than a few large frames?

Mike

[Giuseppe Larosa · Accepted solution]

Hello Mike,

yes the presence of IFG is less important for the bigger frames in percentage.

However, when we say full rate we all agree what is possible taking in account IFG because it is part of the standard.

Hope to help

Giuseppe

[Joseph W. Doherty]

"I am just wondering what the actual throughput rates are that you get on various Ethernet links (10Mpbs, 100Mpbs, 1Gig)

I am not talking about the amount of data in a packet or frame, just raw bitrate... "

They would have 10, 100, 1,000 Mbps, etc., "raw". However . . .

"I guess then that frame size would have an affect on the throughput of data as lots of little frames would require more IFGs than a few large frames? "

Yes, and not just IFG, as Giuseppe noted in his first post, there's other "overhead" that very much impacts the effective throughput of "data" based on the ratio between the two.

The issue of overhead, such as Ethernet frame header (including optionally VLAN tags), also applies to other overhead, such as protocol, e.g. IP (20 bytes) and TCP (20 bytes, plus options), and other media too, when finding effective data throughput vs. media bandwidth.

Consider sending 1 character per IP TCP packet using telnet. If you push the media at maximum, your effective data rate is rather low. (NB: one reason for Naggle's algorithm.)

To increase effective data transfer rate, larger packets are so helpful, it's one of the reasons for "jumbo" Etherent.

More info can be read here: http://sd.wareonearth.com/~phil/net/overhead/

[mike_guy29]

Thanks for that. Very helpful

Mike

[Pavel Bykov]

Cisco has written on this subject and explains the calculations:

http://www.cisco.com/web/about/security/intelligence/network_performance_metrics.html