Advanced Operating Systems Project

Requirements

1. Library Dependencies

   liburing
   

2. Programs should have CAP_SYS_NICE capability

   sudo setcap 'cap_sys_nice=eip' fcp
   sudo setcap 'cap_sys_nice=eip' fcp2
   

Build

mkdir build; cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make

Run

The interface is similar to GNU cp. For eg., to make a copy of dir1 called copy_dir1

./fcp -r dir1/ /path/to/copy_dir1/

Features / Flags:

Flag	Description	cp	fcp
-h	Prints help	✓	✓
-r	copy recursively	✓	✓
-b B	total buffer size in KiB (default: 128KiB)	✓	✓
-n N	# of sub-buffers to use (default: 2)		✓
-k	(io_uring) use a separate kernel thread to poll SQ (default: false)		✓
-q Q	size of SQ (default: 16384)		✓

Benchmarks & Tests

All benchmark scripts and tests can be found inside the tests/ folder.

1. sanity_checks.sh: Runs some basic sanity checks by copying files/folders and verifying correctness via diff:

   ./sanity_checks  ../build/fcp

2. benchmark_single.sh: Benchmarks the time taken to copy a single file

   sudo ./benchmark.sh <size-in-MB> <cp-executable> [opts] 

3. benchmark_multi.sh: Benchmarks the time taken to copy a folder with multiple files

   sudo ./benchmark.sh <size-in-MB> <num-files> <cp-executable> [opts] 

4. To run a test: Choose a test config file from tests/graphs/test_configs/ directory.

   python run_test.py -f <path_to_config_file> -r <result_directory_path> -t <target_directory_to_test> --bin <path_to_build_directory>
   

5. To run all tests: Go to tests/graphs/test_configs/ directory.

   ./runall.sh
   

Results

To generate results:

1. First run desired tests as described in step 4 or 5 above.
2. Use the jupyter notebook files (.ipynb) from results_ssd_1k_cc/ directory to generate the results. Use the corresponding results (with matching name prefix) to generate the graphs.