keomabrun / k7

Connectivity Traces on Tape

k7

CCBY: Cassette by Alvaro Cabrera from the Noun Project

Usage

Check file format

python -m k7.k7 --check myfile.k7

k7 format

{"location": "grenoble", "tx_length": 100, "start_date": "2018-01-11 16:32:22", "stop_date": "2018-01-13 16:21:30", "node_count": 44, "channels": [11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26], "interframe_duration": 100}
datetime,src,dst,channel,mean_rssi,pdr,tx_count
2018-01-11 16:33:07,18,19,11,-53.2,1.0
2018-01-11 16:33:07,17,14,11,-84.03,0.97
2018-01-11 16:33:07,23,27,11,-83.88,1.0
2018-01-11 16:33:30,16,14,11,-67.03,1.0
2018-01-11 16:33:30,22,49,11,-70.0,1.0
...

Header

Each k7 starts with a one-line header. The header is the json dump of a dict. The header contains the dataset meta data. Ex:

{"location": "grenoble", "tx_length": 100, "start_date": "2017-06-20 16:22:15", "stop_date": "2017-06-21 10:29:29", "node_count": 50, "channels": [11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26], "interframe_duration": 10}

Data

datetime	src	dst	channel	mean_rssi	pdr	tx_count
iso8601 string	int	int	int	float	float (0-1)	int

Standard example:

datetime	src	dst	channel	mean_rssi	pdr	tx_count
2017-12-19 21:35:41	0	1	11	-74.5	1.0	100

The source or destination can be empty (i.e when measured on all the neighbors of the src):

datetime	src	dst	channel	mean_rssi	pdr	tx_count
2017-12-19 21:35:41			11	-74.5	0.7	100

Unknown channel:

datetime	src	dst	channel	mean_rssi	pdr	tx_count
2017-12-19 21:35:41	1	2		-79.5	1.0	100

When the channel value is empty, it means that the channel is unknown for that measurement.

k7's Issues

add utils to convert mercator or smip to k7

site -> location

Samples in README.md have some errors

1. tx_count is missing in data lines (link)
2. start_date and end_date are not iso8601 strings (link)
3. datetime is not an iso8601 string (link1, link2, link3)

Related discussions / issue tickets:

• keomabrun/dense_connectivity_datasets#3 (comment)
• #7

Make channel back into a unique integer value

The current channels column requires extra parsing function and is not useful in practice.
The channel is either known or unknown.

For instance:

• in Mercartor, the channel in know so the value with be for instance 11
• in SMIP, the channel value in unknown so the value will be empty (no value between to commas)

This new format does not allow to group multiple channels (when multiple channels have the same RSSI value for instance) but that is rarely the case and the grouping (compression) does not bring a huge benefice.

If the channel is unknown, then the script that use the trace (e.g the 6tisch simulator) can set all the channels to the same value.

The channels that were used during the trace collection are supposed to be provided in the JSON header.

This reverts #2

Add script to convert SMIP into K7

As SMIP only reports aggregated traces, a conversion with basic rules need to be applied in order to convert traces into K7.
Before assuming a link is not available (i.e PDR=0), we should wait for 3 HR.

From LD:

The discovery frame is at least 256*256 slots long, that is 475.136s.
22.5 mins is 1350s.
Ceil(1350/475.136) = 3.
So my expectation is that we could hear a max of 3 discovery packets per HR under regular circumstances.
The only time this could be exceeded that I can think of is if we are unable to generate an HR (due to e.g. queue congestion), then the counters are not cleared and we try to generate the HR one interval later.

channel -> channels

• "11" -> (11,)
• "11-25" -> (11,12,13,14,15,16,17,18,19,20,21,22,23,24,25)

Remove fill function

K7 should not be a matrix. That choice should be left to the script that use the trace.

Check date format

The --check feature should check if the datetime format is the right one.