Open Djent- opened 2 years ago
git-checkout
the djent--preprocess
branch to play with preprocessing. It does not help. Neo4j is constant time. As Python collects more and more data, it takes longer to search and insert. From the start, it processes packets slower (~44 packets/sec) and is then down to ~15 packets/sec after 4,000 packets processed.
python3 NetFrenzy.py -c 5k.pcap --count 5000 --preprocess
Time for 5,000 packets: 1:44 on main
, 4:01 on djent--preprocess
A last ditch effort may be to use dictionaries instead of lists but I will leave that as an exercise to the reader @tvldz @broosa
I recommend focusing on the other enhancement issues
This is an attempt to trade Python resource usage (RAM) for Neo4j
MERGE
query time. I will take a first stab at thisThe goal is to maintain the same amount of data as the current ingestion technique