probabilistic-packet-filter

Project Bi-Weekly Update: Enhanced Memory & Time Benchmarking with HashSet vs BloomFilter

Student: Jonathan Ami
Date: April 4, 2025

---

Planned Activities:

• Create new benchmarking scripts to isolate and profile memory and time usage
• Chart results from proc_benchmark.sh and insertion_time_benchmark.sh
• Modify packet filter code to simulate client server architecture with a working packet filter

---

Progress Update:

New Testing Methods:

Two new benchmarking tools were introduced:

1. insertion_time_benchmark.sh: Runs time-based insertion tests from 100k to 5M entries using both data structures, writing results to CSV.
2. proc_benchmark.sh: Profiles /proc/<pid>/status data after 1-second runs to capture memory usage for Bloom Filter and HashSet, exporting to bloomfilter_mem.csv and hashset_mem.csv.

---

Memory Benchmark Results:

VmRSS (Resident Memory) showed expected trends as insertions scaled:

N	VmRSS (KB) - HashSet	VmRSS (KB) - BloomFilter
100k	3244	2740
1M	20428	3680
2M	38976	4932

Observations:

• At 100k insertions, memory usage is relatively close.
• At 1M, the HashSet uses over 5.5x the memory of the Bloom Filter.
• By 2M insertions, that difference increases to nearly 8x, confirming the Bloom Filter’s substantial memory efficiency.

---

Insertion Time Benchmark Results:

Insertion timing, recorded using the insertion_time_benchmark.sh, also showed favorable results for BloomFilter:

N	Time (ms) - HashSet	Time (ms) - BloomFilter
100k	~1.47	~2.69
5M	~81.10	~182.17

Analysis:

• BloomFilter is approximetly 2x slower than the hashset
• Likely cuased by multiple hashfunctions and bit manipultaion overhead

---

Repository Changes:

• Added benchmark_results/ folder for all generated CSV data.
• New .sh scripts to handle batch timing and /proc memory capture.
• Two new rust projects created to benchmark performance
• Both Bloom and HashSet test projects now support CLI args for insertion count and FP rate.
• Updated the pkt filter code to simulate client server architecure

---

Next Steps:

• Full writen report + demo of packet filter