Register (or login) on our website and you will not see this ad.
|
|
Maximum TCP/IP throughput is 96.8% of the sync rate.
I do believe these are now no longer in use under G.INP enabled. 96.8% are probably less now.
Before G.INP
BT IP 77.44Mb
After G.INP
BT IP 77.35Mb
Edited by adslmax (Wed 01-Apr-15 15:15:00)
|
|
|
|
I haven't checked your numbers, but you're right about the scale of overhead for IP and TCP headers over Ethernet LAN.
However, the bulk of the overhead between router and ISP is encapsulated in the "IP profile".
The difference between a sync speed of 40,000 and the corresponding IP profile of 38,717 is an indication of the volume of IP data that can be transferred over the DSL link; essentially, 3.3% is the overhead of getting your IP packets from the router to the ISP (including those extra 8 bytes in PPPoE).
With an IP profile of 38717, the throughput of "a download" (ie just the payload data) is around 37.5Mbps at absolute best. This reduction is accounted for by the headers in IP and TCP.
On an 80/20 connection, the IP profile is 77.4Mbps, and the top end throughput during a download around 75Mbps.
Note: G.INP reduces the IP profile slightly (0.1Mbps).
|
|
|
Here are some stats from my Zen connection. First column is BRAS speed, second is SYNC (line) speed and then BRAS/SYNC. Average at the end as a %age.
| Text | 1
23
45
67
89
10 | 64387 66519 0.967949
58873 60823 0.9679457900 59818 0.967936
59714 61692 0.96793760056 62045 0.967943
59745 61724 0.96793861729 63773 0.967949
60293 62289 0.96795660643 62651 0.967949
Average: 96.7944 % |
Edited by deleted (Wed 01-Apr-15 15:48:18)
|
|
Register (or login) on our website and you will not see this ad.
|
|
|
I halved it from the throughput on the 80/20 service.
|
|
|
Note: G.INP reduces the IP profile slightly (0.1Mbps). Are you sure about that?
Since attaining exactly 60,000kbps I've had a 58.01Mbps IP Profile. Rather than the 58.07Mbs I would expect from the 0.9679 factor. Download speedachieved during the test was - 51.52 Mbps
For your connection, the acceptable range of speeds is 40 Mbps-58.01 Mbps .
Additional Information:
IP Profile for your line is - 58.01 Mbps # xdslcmd info --pbParams
xdslcmd: ADSL driver and PHY status
Status: Showtime
Retrain Reason: 0
Last initialization procedure status: 0
Max: Upstream rate = 16964 Kbps, Downstream rate = 58800 Kbps
Bearer: 0, Upstream rate = 16961 Kbps, Downstream rate = 60000 Kbps
Bearer: 1, Upstream rate = 0 Kbps, Downstream rate = 0 Kbps
Discovery Phase (Initial) Band Plan
US: (7,32) (871,1205) (1972,2782)
DS: (33,859) (1216,1961) (2793,3970)
Medley Phase (Final) Band Plan
US: (7,32) (871,1205) (1972,2782)
DS: (33,859) (1216,1961) (2793,3970)
|
|
|
|
77.44 -> 77.35 here.
|
|
|
|
|
|
|
Yes. The overhead is essentially down to overheads on the framing, IP and TCP overheads.
The usual maximum packet size is 1500 bytes (there are so-called "jumbo-packets", but they are only used for specialist purposes, like local storage over IP and require special support). However, within that 1500 byte packet there is a 20 byte IP packet header and a 32 byte TCP header. That gives you a payload of 1448 bytes. However, there is also an Ethernet framing overhead comprising another 38 bytes. So, for every 1448 bytes of data there are 1538 bytes sent over the link. That's 94.1% payload efficiency, which means for a 40mbps sync there is a maximum of 37.6mbps data throughput available.
The proportion is the same no matter what the sync speed. So an 80mbps will have a maximum TCP payload throughput of 75.2mbps.
*** edited as there was a slight mistake where I'd used 1440 as the payload size ***
Note that this is for IPV4. IPV6 has larger overheads.
Also, there is an overhead for "ack" packets in the opposite direction (About 4%, so downloading at 40mbps requires a minimum upload bandwidth of 1.6mbps).
UDP protocols have different overheads, but that's not normally used for speed tests.
Ethernet frame headers are 14 bytes + 4 more if using 802.1Q + 2 byte checksum, TCP headers are 20. The overheads are 18 bytes for Ethernet, 8 for PPP, and 40 for TCP/IP = 68 bytes per frame.
If the modem and PPPoE router support baby jumbo frames you get 1460 bytes of payload per 1528 bytes on the wire or 1452 bytes per 1520 on the wire if not.
EDIT: Forgot the VLAN tags - extra 4 bytes, and checksum, another 2.
Also the acknowledgement overhead is very unlikely to be 4% as virtually nothing acknowledges every frame and if it did download speeds would be tedious over any kind of latency, selective and cumulative acknowledgements are used to reduce round trip and effect of loss.
Edited by deleted (Wed 01-Apr-15 17:06:32)
|
|
|
Unchanged sync?
79,987 -> 79,999, it says.
How about these settings?
before:
Bearer 0
INP: 0.00 0.00
INPRein: 0.00 0.00
delay: 0 0
PER: 1.65 3.98
OR: 116.09 64.22
AgR: 80103.09 20063.54
After G.INP:
Bearer 0
INP: 46.00 47.00
INPRein: 0.00 0.00
delay: 0 0
PER: 0.00 0.00
OR: 0.01 0.01
AgR: 80614.82 20102.08
Bearer 1
INP: 4.00 4.00
INPRein: 4.00 4.00
delay: 3 0
PER: 16.06 16.06
OR: 95.62 31.87
AgR: 95.62 31.87
I *think* "AgR" means aggregate rate, and includes some of the overhead bits, as well as the main "throughput" we want from the link.
What we see as "sync speed" should be a combination of the throughput on both bearer 0 and bearer 1. Bearer 1 is close to using 0.1Mbps. But bearer 0 has gone up from the original... so it depends what the other overheads are.
|
|
|
Wow so a 0.1 or 0.2% difference. I'll go throw a party
|
|
The author of the above post is a thinkbroadband staff member. It may not constitute an official statement on behalf of thinkbroadband.
|
|
|
Pages in this thread: 
Print Thread
adslmax
