CS-331-Benchmarking-Client-Server-Architectures

Single-Process Server: Handles one client at a time.
Multi-Process Server: Spawns a new process for each client.
Multi-Threaded Server: Uses threads to handle multiple clients concurrently.

Running the Servers

Start the desired server in a separate terminal:

python3 single_process_server.py
python3 multi_process_server.py
python3 multi_threaded_server.py

Running the Benchmark

Execute the benchmarking script to test each server implementation with different client loads:

./finalBenchmark.sh

This script:

• Runs each server with a range of concurrent clients (10 to 100).
• Records execution time (latency) for each test.
• Generates performance plots.

Alternatively, you can manually test a specific client load:

./start.sh client.py 10

Performance Evaluation

After running the benchmarks, execute:

python3 plot_results.py

This script generates and saves plots comparing the performance of all three server types.

Expected Results

• Single-Process Server: Latency increases linearly as clients wait sequentially.
• Multi-Process Server: Can handle multiple clients concurrently but has higher overhead because new processes are created.
• Multi-Threaded Server: Best scalability due to lightweight threads.

Results Visualization

Clients	Single-Process Server (s)	Multi-Process Server (s)	Multi-Threaded Server (s)
10	30.19703	3.37562	3.18338
20	60.13437	3.43206	3.26826
30	90.13987	3.57823	3.38463
40	120.21113	3.69099	3.49122
50	150.23126	4.18303	3.83011
60	180.65150	4.21820	3.85501
70	210.52304	4.31251	3.88819
80	240.71175	4.55438	3.95132
90	270.95691	4.72590	4.10546
100	301.04635	4.82261	4.18694

Execution Time vs. Number of Clients

Comparison of Multi-Process and Multi-Threaded Servers

Conclusion

This benchmarking experiment demonstrates that while all three server designs function correctly:

• The Single-Process Server is inefficient for high concurrency.
• The Multi-Process Server improves performance but incurs process creation overhead.
• The Multi-Threaded Server offers the best performance and scalability for concurrent requests.

---

Author: Bhavik Patel - 22110047
Institution: IIT Gandhinagar