TCP Prague Classic ECN AQM Fallback v2 - Classic Rate Ratio Heat-Map Home Evaluation Results
Classic Rate Ratio: Heat-Maps Detection Success: Heat-Maps Show all
Internals
Algorithm Experiment Setup Implementation
Further Information

TCP Prague Classic ECN AQM Fallback v2 - Classic Rate Ratio Heat-Map

The Classic Rate Ratio (CRR) is the percentage that average Cubic flow rate reduces to when competing against Prague flows, compared to when competing against equivalent Cubic flows.

More precisely, CRR = xCC / xCP *100%, where xCP is the average rate of long-running Cubic flows competing against Prague, and xCC is the average rate of long-running Cubic flows competing against Cubic flows that replace Prague flows in otherwise exactly the same traffic scenario.

How to Read the CRR (Classic Rate Ratio) Matrix

The key below indexes the smallest 4x5 matrices that each represent 20 experiments with different combinations of traffic flows (although only 9 are actually used for CRR matrices; the rest are left white).

Flows

Each traffic scenario is denoted by two descriptors placed either side of a colon, with the Prague flows always given first.
n
denotes n long-running flows, where n is a number, typically 0, 1, 9
L
denotes a
Low loadExponential arrival process, averaging 1 request per second for the 4Mb/s link rate, scaled pro rata for faster links
of
short web-like flowsFlow sizes Pareto distributed with α = 0.9; min 1KB, max 1MB.
1L
denotes '1' and 'L'
Example:
9:1L
denotes 9 Prague and 1 Cubic-ECN long running flow plus a low load of short web-like Cubic-ECN flows. This is the traffic scenario at the intersection of row 9: and column :1L.
The scenarios with solely short flows in either class (L:* and *:L) are not used for 'fairness' experiments, because a short flow does not have a steady-state rate, so there is nothing to compare. However, the columns where these scenarios would be are kept, but coloured white, so that it is easier to visually compare with the 4x5 matrices used in other experiments, such as AQM detection. Nonetheless, the scenarios with short and long flows are used, in order to guage the background effect that short flows have on how long-running Prague and Cubic flows interact.
Flows↓→   :0   :1   :9   :L   :1L ←Cubic‑ECN
1:
9:
L:
1L:
Prague↑

Each of these 9 experiments has been run with 5x5=20 different combinations of link rate and base RTT, and in turn with different combinations of code, of traffic and of AQMs. The heat-maps give an intuitive feel for whether there are problem areas, and if so what their common factors are.

Click 4x5 matrix To dive into detailed time-series plots tracking all the relevant metrics, click on one of the 4x5 matrices.

Click (5x5)x(4x5) matrix To view more detailed Classic Rate Ratio heat-maps (with the actual percentages written in each cell) as well as whisker plots of normalized rates and Detection RAG matrices for any one combination of AQM and host software (5x5)x(4x5)=500 experiments click on the wide margin of a matrix (highlights in red when mouse over).
Classic Rate RatioThe percentage that average Cubic flow rate reduces to when competing against Prague flows, compared to when competing against equivalent Cubic flows.
(CRR)
Legend:
200 141 100 71 50 35 25 18 13 9 6 4 3 2 2 1
Code
Algorithm
 Fallback v2.2 (detection only) Fallback v2.2 Fallback v2.2
RTT indep?
N N N
Traffic Staggered?
N N Y
Mixed RTTs?
N N N


↓ Bitrate (Mbps) : Base RTT (ms) → 5 10 20 50 100 5 10 20 50 100 5 10 20 50 100
One AQM CoDel 4
12
40
120
200
COBALT 4

12
40
120
200
DualPI2 4

12
40
120
200

Commentary


Bob Briscoe and Asad Sajjad Ahmed
©bobbriscoe.net Ltd 2019-2020