Re: [tcpm] [tsvwg] "David's proposal" for cc class terms (was RE: L4S status tracking)

[Jonathan Morton <[e-mail]>]

> On 13 Nov, 2019, at 6:15 pm, Black, David <[e-mail]> wrote:
> 
> We are dealing with two classes of congestion controls.  For lack of better terms the following class names are based on what the transport protocol throughput is proportional to where ‘p’ is the loss and/or congestion marking probability:
>                - 1/sqrt(p)-class congestion controls: Includes most existing TCP congestion control algorithms, e.g., NewReno, CUBIC.
>                - 1/p-class congestion controls: Includes DCTCP congestion control.
> Keep in mind that p is a probability that is usually << 1 when expressed as a decimal, e.g., p=0.01 represents a 1% loss/marking rate.

Specifically CE marking or loss.  SCE marking can orthogonally elicit a DCTCP-style 1/p-class or an SCE-specific 1/(p*sqrt(p)) class of response, overlaid on the normal 1/sqrt(p) response to CE.

To give these a more intuitively recognisable face, may I suggest:

- "MD" for Multiplicative Decrease, for the traditional 1/sqrt(p)-class TCP response.

- "DCTCP" for the 1/p-class response introduced by that scheme.

- "ELR" for Explicit Load Regulation, for SCE's preferred 1/(p*sqrt(p))-class response.

ELR was an early name considered for SCE, using a control element commonly found in diesel locomotives (known as a Load Regulator) as inspiration.  This generally has a slow adjustment mode for small disturbances from steady state, and a fast adjustment mode for larger disturbances.  The MD response corresponds to the fast adjustment mode, and the ELR response with the traditional linear cwnd growth correspond to the slow adjustment mode.  Since SCE refers specifically to the signalling method, we can use ELR to avoid confusion when referring to the response function.

Incidentally, another way of writing the above formulae is as an exponent to p:

- p^-0.5 is 1/sqrt(p)

- p^-1 is 1/p

- p^-1.5 is 1/(p*sqrt(p))

The above may also be more readable, if not as familiar to non-mathematicians.

 - Jonathan Morton