Trace_Complexity_Map

A basic exmple with a set of function that generate a trace complexity map for two traces. Based on the paper: On the complexity of traffic traces and implications.

This version of the code was used in Harnessing Generative Pre-Trained Transformer for Datacenter Packet Trace Generation However, the method accurately reflects the original paper.

Traces:

This comes with two HPC example papers.

hpc_cesar_mocfe.npy
hpc_exact_boxlib_cns_nospec_large.npy

More information on the traces can be found in the complexity paper.

Usage:

Install dependencies. With all trace files in the root directory, run main. This will produce a complexity map for the two provided traces

New traces must follow the format and datatype of the given examples. Each trace is:

• A two column ndarray of integers.
• Integers represent node IDs in a network.
• It is assumed that IDs are consecutive.
• There are n IDs
• The max ID is n-1
• The integers range is [0,n-1].

Inside the code, the following function will calculate complexity for the a trace 'trace'

from complex_func import *
trace_c_dic = get_trace_complexity(trace, number_of_unique_IDs=n ,comp_types='Main')

Produces a dict with results for temporal and non-temporal complexity types. It's fields are e.g.,:

{'temporal': 1,
 'non_temporal': 1,
 'total': 1,
 'entropy':1,
 'pair_entropy': 1}

Further documentations is inline with the code.

Result

After running main the resulting figure should be the following: