cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
2488
Views
0
Helpful
4
Replies

Total bandwidth utilization on a full-duplex port

pweinhold
Level 1
Level 1

My co-worker and I are having a disgreement over this question and perhaps someone can help us settle it.  Here's the question:

On a 100 Mbps full-duplex port, is the total, aggregated bandwidth 100 Mbps or 200 Mbps?  I think it's 200 Mbps: 100 Mbps TX and 100 Mbps RX.  For example, see the following output from a full-duplex, 100 Mbps interface:

5 minute input rate 4332000 bits/sec, 536 packets/sec

5 minute output rate 31979000 bits/sec, 3100 packets/sec

On this interfaces, the RX (input) channel shows about 4.3 Mbps, while the TX (output) channel shows about 31.9 Mbps.  According to my aggregated 200 Mbps theory, the RX is only 4.3% utilized and the TX is 31.9% utilized.  My co-worker, on the other hand, says that he adds the two figures up to determine interface utilization, so for this example, he would consider this interface about 36% utilized.

I think I'm correct, because after all, isn't that the whole idea of full-duplex, that you get bi-directional throughput?  I just need some verification so I can throw it back in his face and declare victory...

Thanks

4 Replies 4

Exactly!  Thanks for the validation.

egilles123
Level 1
Level 1

What was the answer. That link is no longer valid

It would be 100 Mbps in either direction, which can be done concurrently.

I.e. 200 Mbps aggregate, but host A can only send 100 Mbps to host B, which it could also do with 100 Mbps half duplex.  The big difference is, with full duplex, host A can send 100 Mbps to host B and host B can send 100 Mbps to host A, but with half duplex, hosts A and B share 100 Mbps. In theory, for the latter, host A might send 75 Mbps to host B while host B sends 25 Mbps to host A, but in practice (assuming the half duplex is Ethernet), collisions reduce effective concurrent effective data transfer rates.  (In the worst possible case, if lots of hosts are sharing the same half duplex link, effectively, they might be unable to send any data between hosts.)

BTW, 4 Mbps Token-Ring was also not full duplex, but it managed sharing bandwidth differently from Ethernet.  The aforementioned Ethernet issue of hosts being unable to effectively transfer data at all, was not a problem in TR.  (NB: I was working in a Fortune 500 company when they begin to transition from 4 Mbps TR to 10 Mbps half duplex Ethernet.  Those segments moved to 10 Mbps half duplex Ethernet "crashed and burned".  They stopped migrations, upgraded network to equipment that supported 10 Mbps full duplex, and then, migration went forward just fine.  [BTW, for, then, day-to-day network host usage, saw little difference between 4 Mbps TR and 10 Mbps full duplex Ethernet.])

PS:

Laugh - for the record, my job had nothing to do with supporting or designing the network.  I.e. I had nothing to do with suggesting moving from 4 Mbps TR to 10 Mbps half duplex Ethernet, was "good".

In fact, in my position, I "suggested" to the departmental manager, if possible, we somehow are just so busy, we might be the last to implement moving our hosts to 10 half.  Further, if additional IT resources might be provided to help us make such a transition, again, we're, at the moment, just so busy, unsure, at the moment, we can meet our deadlines with any interruptions.  As much as we too would "benefit" from such an "upgrade", perhaps other departments that want the "better" network, can take advantage of any additional IT resources to make their transition before us.  So, thank you very much, and we get back to you on that ASAP.

Laugh - we "unfortnately" missed taking advantage of 10 half.  We only found time to make the migration after network supported 10 full.  ; )

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: