ADSL overhead

Yesterday I’ve described the difference between line rate and bit rate (actually physical layer gross bit rate and physical layer net bit rate). Going to the other extreme, we can measure goodput (application-level throughput), which obviously depends on multiple factors, including the TCP window sizes and end-to-end delays. There are numerous tools to test the goodput from/to various locations throughout the world (speedtest.net worked quite nicely for me) and you’ll soon discover that the goodput on your DSL line differs significantly from what the ISP is advertising.

Apart from obvious marketing ploys, the difference arises due to the baroque Russian dolls encapsulation stack used on ADSL (see the ADSL-based protocol stacks section in this document). A typical PPPoE connection runs IP over PPP (2 bytes) over PPPoE framing (6 bytes) over Ethernet framing (14 bytes) over RFC 1483 bridged connection (8 bytes) over AAL5 (8 bytes) over ATM cells (5 byte overhead for every 48 bytes of payload). Every IP packet is thus burdened with 38 bytes of DSL protocol stack overhead plus 10.4% ATM cell tax. Assuming the average IP packet size in the Internet is around 500 bytes, the DSL protocol stack overhead is 6%, resulting in a total of over 16% of lost bandwidth just to accommodate the committees that constructed this monster.

Update @ 2009-03-27: I forgot yet another encapsulation technique: AAL5. I also need to investigate whether the RFC 1483 bridged encapsulation really uses the 8-byte header (there are numerous options in the RFC).

10 comments:

  1. I am amazed to understand why windows vista is not able to get the upload and download speed more than 2 Mbps. I made all the changes but no effect. Everytime speedguide shows the same output.

    regards
    shivlu jain
  2. You need a server with fast uplink that is close to you. Testing with a far-away server will always result in low throughput due to Bandwidth-delay product.
  3. It's for this reason that the best MTU to use for an ADSL connection is 1454 over PPPoE, and 1462 over PPPoA.

    http://whirlpool.net.au/wiki/?tag=ADSL_Theory_PPP

    Further improvements can be made by tweaking RWIN, though good for a predictable VPN environment, this is fairly useless over the internet, as ideal RWIN value varies greatly, depending on the latency to the host you're receiving from, their upstream, and your provider's upstream, which is likely to be contended at least some of the time.
  4. @Anonymous: Somehow my calculations of total overhead disagree with the link you've quoted by 4 bytes (so my "ideal" MTU is 1450 for PPPoE).

    Anyhow, the differences are negligible. The ratio between "goodput" (application data) and raw bits on the wire is 85.82% for "optimal" MTU and 85,61% for MTU=1492.

    Also note that a lot of packets are 40 bytes long (TCP ACK), where the overhead is a whooping 62,6%, so the debate about the "optimal" MTU is somewhat academic.
  5. I think in a typical PPPoE ADSL connection we don't use RFC1483 bridging, but we use AAL5 SNAP encapsulation (also specified in RFC1483) with 8-byte header and 8-byte trailer. There is also a variable padding. I found this information here:

    Measuring the Utilization of ATM PVCs

    http://www.cisco.com/en/US/tech/tk648/tk362/technologies_tech_note09186a0080093c9a.shtml
  6. Hmmm ... time to find someone with a protocol analyser ...
  7. Hello
    we here use business class ADSL link with RFC1483 protocol... directly routed links without any ppp negotiations.
    We assume having zero overhead then?
  8. You still have to deal with RFC1483 encapsulation, AAL5 encapsulation and ATM cell tax (5 bytes of overhead in each 48-byte cell + some unused space in the last cell of the packet). Here's a good explanation of what you have:

    http://cisco.biz/en/US/tech/tk648/tk362/technologies_tech_note09186a0080093c9a.shtml
  9. You neglected the 4 byte Frame Check Sequence at the end of an ethernet frame. Therefore ethernet over head is 18 bytes not 14. That makes total overhead on PPPoE with LLC 42 bytes not 38. The optimal TCP/IP MTU is thus 1446 to fit into 31 ATM cells with no padding.

    PPPoA VC-MUX
    ATM 31*48 Full Cells 1488
    AAL5 Trailer -8
    PPP -2
    Optimal Layer 3 MTU 1478

    PPPoA LLC
    ATM 31*48 Full Cells 1488
    AAL5 Trailer -8
    PPP -2
    LLC Header -8
    Optimal Layer 3 MTU 1470

    PPPoE LLC
    ATM 31*48 Full Cells 1488
    AAL5 Trailer -8
    PPP -2
    PPPoE Header -6
    Ethernet Header+FCS -18
    LLC Header -8
    Optimal Layer 3 MTU 1446
    Replies
    1. I don't think Ethernet FCS is included in the ATM payload, but then I may be wrong (wouldn't be the first time).
Add comment
Sidebar