Ethernet utilization formula per RFC2819

Ugrgh. Someone cancelled out a factor of ten.

rewrite as:

Pkts * (96 + 64) + (Octets * 8) Utilization = ------------------------------------- Interval * 100,000

96 is the number of bit times (minimum) for the interframe gap. 64 is the number of bits in the preamble + SFD. (The octet count doesn't include these, so we have to add them in for each packet.) 8 is the number of bits in each octet.

Utilization will be in %, so rewrite as:

Pkts * (96 + 64) + (Octets * 8) Utilization = ------------------------------------- * 100% Interval * 10,000,000

The 10000000 in the denominator is the raw bit rate, in bits per second.

In other words, Utilization (sic) is the number of bits transmitted, divided by the number of bits that could have been transmitted, multiplied by 100%.

RFC2819 lists its author as Steve Waldbusser. Perhaps you'd like to contact him and ask him why he chose to make this hard to understand. There could be a valid reason!

Regards, Allan

Thomas Bahls wrote in part:

Because there's a certain amount of overhead per packet (src & dest MAC, inter frame gap, etc). A network might be very heavily loaded if all the data is in very small packets.

As the other response, someone has divided top & bottom by ten. Octets are 8 bits. 96 + 64 would correspond to packet overheads, IIRC 96 for preamble & MACs, 64 for IFG.

-- Robert

This makes it quite clear, so clear that one wonders why I didn't get it myself. I apparently overlooked the fact that we transfer more than the raw octets starting with dstMAC. The point was nicely disguised by one of those tricks mathematicians seems to enjoy: They almost 'randomly' divide and expand entire equations by 'magic' factors. We loved that at university, kind of. ;-)

I think I'd better not bore Steve with this now I see what the parts of the equation represent. Instead, I will carefully keep a copy of your explanation and enjoy the understanding.

Thanks for your prompt f'up (also to Robert), /Thomas.

This is for 10 mbps Ethernet. For 100 and 1000 mbps Ethernet do I just divide by 100 and 1000? Since they use different encoding schemes something tells me it is not that simple.

No, it is that simple. 10Gbps Ethernet also follows the same rule. The line encodings are different (and have different overhead) but that is made up for by changing the symbol rate on the line to make the effective usable bit rate 10M, 100M, 1G, 10G, etc. The IFGs and preambles also stay the same (nominal) number of bits.

Regards, Allan

Not counting frame extension for half duplex gigabit ethernet.

And maybe someday the useless preamble on continuous signaling systems will be removed.

-- glen