Dataset Analyzer

PTP Analyser

usage: analyze [-h] -f FILE [--analyze-only]
               [--disable {ewma,sample-average,sample-min,sample-max,sample-mode,sample-median,ls,kalman} [{ewma,sample-average,sample-min,sample-max,sample-mode,sample-median,ls,kalman} ...]]
               [--no-optimizer] [--no-cache] [--infer-secs]
               [--batch-size BATCH_SIZE] [--skip SKIP] [-N NUM_ITER]
               [-t TIME_INTERVAL] [--prefix PREFIX] [--verbose]
               [--pkts-no-drift-comp] [--bias {pre,post,both,none}]
               [--optimizer-fine] [--optimizer-force]
               [--optimizer-metric {max-te,mse}]
               [--optimizer-max-window OPTIMIZER_MAX_WINDOW]
               [--optimizer-no-stop]
               [--drift-est-strategy {loop,unbiased-two-way,unbiased-one-way,unbiased-one-way-reversed}]
               [--drift-est-pkts {sample-min,sample-max,None}]
               [--drift-est-loss {max-error,mse}] [--eps] [--dpi DPI]
               [--latex] [--cal-quantum CAL_QUANTUM]
               [--cal-decimation CAL_DECIMATION] [--cal-drift-comp]
               [--cal-prob-thresh CAL_PROB_THRESH CAL_PROB_THRESH]
               [--cal-static-thresh]

Named Arguments

-f, --file

Dataset file to analyze.

--analyze-only

Run analyzer plots only and don’t run any post-processing. The analyzer will then process only the data that is already available in the dataset and the data produced by the dataset reader, such as two-way time offset and delay measurements. Useful to inspect results without much wait.

Default: False

--disable

Possible choices: ewma, sample-average, sample-min, sample-max, sample-mode, sample-median, ls, kalman

Algorithms to disable

Default: []

--no-optimizer

Whether to optimize estimators’ parameters.

Default: False

--no-cache

Whether to disable caching of optimal configurations

Default: False

--infer-secs

Infer seconds rather than using the seconds that were actually captured.

Default: False

--batch-size

Maximum number of observation windows processed at once on window-based algorithms.

Default: 4096

--skip

Fraction of the dataset to skip on analysis to ignore transient effects.

Default: 0.2

-N, --num-iter

Restrict number of iterations.

Default: 0

-t, --time-interval

Specific time interval to observe given as “start:end” in hours.

--prefix

Prefix to prepend to saved plot and cache files.

--verbose, -v

Verbosity (logging) level

Default: 1

Packet Selection/Filtering Options

--pkts-no-drift-comp

Whether to disable the drift compensation step applied within packet selection algorithms.

Default: False

Bias Correction Options

--bias

Possible choices: pre, post, both, none

Compensate the bias prior to any post-processing (pre), after post-processing (post), both pre and post post-processing (both) or disable it (‘none’).

Default: “both”

Window Optimizer Options

--optimizer-fine

Whether to enable fine pass optimization (only available for window-based estimators).

Default: False

--optimizer-force

Force optimizer processing even if already done previously.

Default: False

--optimizer-metric

Possible choices: max-te, mse

Estimation error metric for parameter tuning.

Default: “max-te”

--optimizer-max-window

Maximum window length that the window optimizer can return for any algorithm.

Default: 8192

--optimizer-no-stop

Do not apply early stopping on the optimization process.

Default: False

Drift Estimation Options

--drift-est-strategy

Possible choices: loop, unbiased-two-way, unbiased-one-way, unbiased-one-way-reversed

Drift estimation strategy. Select “loop” to use the drift estimates produced by the TLL PI loop or “unbiased” to use the conventional unbiased frequency offset estimator based on intervals measured at the slave and the master. Select specifically which unbiased estimation formulation to use: two-way, one-way, or reversed one-way.

Default: “loop”

--drift-est-pkts

Possible choices: sample-min, sample-max, None

Apply packet selection pre-filtering before frequency offset and drift estimations. Valid when using the unbiased frequency offset estimation strategy.

--drift-est-loss

Possible choices: max-error, mse

Loss function used to optimize the drift estimator.

Default: “max-error”

Plot Options

--eps

Whether to save images in .eps format.

Default: False

--dpi

Images resolution in dots per inch.

Default: 300

--latex

Render plots using LaTeX.

Default: False

Delay Asymmetry Calibration Options

--cal-quantum

Quantization step.

Default: 50

--cal-decimation

Decimation ratio. A value of 1 means decimation is disabled.

Default: 1

--cal-drift-comp

Whether to apply drift compensation on the delta_d estimates used for calibration. Useful when the drift affecting delta_d is expected to be larger than the drift estimation error.

Default: False

--cal-prob-thresh

Probability thresholds.

Default: [0.001, 0.001]

--cal-static-thresh

Whether to disable automatic threshold tuning.

Default: False