Network-Lab

CN lab

CSE/PC/B/S/314 — Assignment 1: Checksum & CRC Error Detection

Client-server implementation matching the "Client — high level view" diagram: create socket → setup server address → connect → read/write data → shutdown.

Files

• common.h / common.c — frame format, 16-bit checksum, generic bit-by-bit CRC (CRC-8/10/16/32 with the exact polynomials from the assignment), the four error-injection routines, and reliable send/recv helpers.
• sender.c — the Sender (client). Reads a file, splits it into Ethernet-sized frames (46–1500 byte payload, padded if short), computes the checksum/CRC over the clean payload, then simulates a transmission error on the copy that is actually sent, and transmits frame header + payload.
• receiver.c — the Receiver (server). Accepts one connection, recomputes the checksum/CRC over the received payload, and compares it to the redundancy field to ACCEPT or REJECT the frame — exactly what a real receiver does. Every frame is logged to a CSV for your report.
• run_evaluation.sh — runs all scheme × CRC-type × error-type combinations automatically and appends everything to results.csv.

Build

make

Run manually

# terminal 1
./receiver <port> [results.csv]

# terminal 2
./sender <server_ip> <port> <file> <scheme> <crc_type> <error_type> [burst_len]

• scheme: checksum | crc
• crc_type: 8 | 10 | 16 | 32 (pass 0 when scheme is checksum)
• error_type: none | single | two | odd | burst
• burst_len: optional, only used when error_type=burst (default 8 bits)

Run the full evaluation automatically

./run_evaluation.sh <file> [host] [base_port]

This produces results.csv with every combination of:

• scheme: checksum, CRC-8, CRC-10, CRC-16, CRC-32
• error type: none, single-bit, two isolated single-bit, odd number of errors, burst

Columns: frame_id, scheme, crc_type, error_type, payload_len, sent_redundancy, recomputed_redundancy, detected, decode_time_ms

Load results.csv in Excel / pandas to build the required tables and graphs (detection rate per error type per scheme, and timing comparisons using the decode_time_ms column plus the total_encode_time_ms the sender prints).

Notes for your report

• Checksum vs CRC: with real data you should see checksum occasionally miss two-isolated-bit or odd-error cases that CRC always catches (CRC's strength is being able to detect all burst errors up to the polynomial's degree and all odd numbers of errors, by construction). Run the evaluation a few times with different random seeds (re-run run_evaluation.sh, it reseeds using time+PID) to build up statistics for the "detected by X but not Y" cases the evaluation section asks for.
• Execution time / resource utilization: the sender prints total/average encode time; the receiver logs per-frame decode time in results.csv. You can additionally wrap runs with /usr/bin/time -v to capture memory/CPU utilization for the report.