longnehc / hypatia

Other
6 stars 2 forks source link

Research and standardize input format for the GNN models #10

Open KamiCreed opened 2 years ago

KamiCreed commented 2 years ago
KamiCreed commented 2 years ago

From @longnehc:

Here is the input data format of the GNN model. You can standardize the measured data into different files as below: - delay.csv: each row contains a five-tuple, the data is standardize in fixed interval (you can set it as 1 second now) source, destination, min. rtt, max. rtt, avg. rtt - bandwidth-tcp.csv: (protocol = tcp) source, destination, avg. bandwidth, avg. transfer bytes - bandwidth-udp.csv: (protocol = udp) source, destination, avg. bandwidth, avg. transfer bytes It would be better to let the interval be a parameter in the script, so we can reformat the data by changing the interval. You can use 0,1,2,...,10 to represent the source (Vancouver) to other AWS regions. If there exists any data such as the jitter or packet drop, you can also prepare a similar file as above and average the data in fixed intervals.