Data Analysis

All files relating to the analysis of data generated by the telemetry.

• CPU_Louvain_Script.py
  • Example Text
• Generate_Subgraphs.py
  • Example Text
• GPU_Script.py
  • Example Text
• GRF_Converter.py
  • Example Text
• MTX_converter_Script.py
  • Example Text
• Output_to_JSON_SCript.py
  • Example Text
• Updated_GPU_Script.Py
  • Example Text

Running the Scripts

If you are running this on your local computer, make sure to change into the Data_Analysis directory of this GitHub before running the command below.

Running Louvain GPU Implementation

To run on a test graph, it must be a graph uploaded to the gunrock directory in .mtx format.

python output_to_json_script.py <algo> <filename.mtx> <exec-type> --ssh-key-path /path/to/your/key.pem --ec2-instance-ip 54.84.25.80 --ec2-instance-username ubuntu

example: .. code-block:: bash

python output_to_json_script.py louvain chesapeake.mtx parallel –ssh-key-path /Users/nani/Desktop/mnKeyPair.pem –ec2-instance-ip 54.84.25.80 –ec2-instance-username ubuntu

Future Plans/Fixes:

• Currently, the full output prints to shell. I want to change this to extract and store only the necessary metric information (time, memory) (Addressed - extracts time metrics)
• I want to redirect output to LaTex/ReadTheDocs instead (Addressed - outputs to JSON)
• I want to parameterize further to incorporate input of any graph file from the local computer instead of simply allowing graphs will already exist in the server to be tested
• I want to integrate/explore more profiling tools to extract all info possible
• I want to integrate a graph viz feature to give a visual representation of the nodes/edges of an input graph. This will be integrated into the LaTeX/ReadTheDocs output solution

Running SGM GPU Implementation (NEEDS TO BE UPDATED)

python3 gpu_script.py sm tri_sm.mtx sequential

Running Louvain CPU Sequential Implementation (NEEDS TO BE UPDATED)

• Must be a graph uploaded to directory in .txt format

python3 cpu_louvain_script.py <filename>

example:

python3 cpu_louvain_script.py chesapeake

Sample Benchmarking Report

GPU Subgraph Matching Parallel on tri_sm.mtx

Average Elapsed Time (ms)	Minimum Elapsed Time (ms)	Maximum Elapsed Time (ms)	Preprocess Time (ms)
0.678062	0.678062	0.678062	26.037

Postprocess Time (ms)	Total Time (ms)	Allocs	Frees	Bytes Allocated
0.273228	27.153015	20034	9888	259571

CPU Louvain Sequential on chesapeake.mtx

Total Duration (ms)	Allocs	Frees	Bytes Allocated
478	179	179	102,993