fs-microbench

This is a fork of a simple filesystem microbenchmark used by Thiago Emmanuel in his file system analysis. See the original code here.

Run

Before running the benchmark, you need to compile the code. Go to src directory and use the following command to compile all code:

make    

To run the benchmark:

./type  [THREADS] [DELAY] [NUMBER_OF_OPERATIONS] [PATH] [BLOCK_SIZE] [DEBUG] >> output.csv

where the type can be sr, sw, rr or rw and the flags:

- THREADS: set the number of threads
- DELAY: represents the interval between requests
- NUMBER_OF_OPERATIONS: set the number of operations
- PATH: is the path of the file that we will do the operations. This file needs to exist before running the command. You should create the file ~/file0 with the required size. 
- BLOCK_SIZE: set the block size
- DEBUG: that must be 'debug' or 'no-debug'

Output

The output is a csv file that contains:

id_thread, id_req, begin_timestamp, end_timestamp, offset, count_bytes

Example

./sr 1 0 5000 ~/file 4096 debug >> output.csv

Example of a sequential read benchmark with 1 thread, no interval between requests, 5000 requests on ~/file0, a block size of 4096 and the debug mode on. This is not a typing error, you should create a ~/file0 before running and reference it as ~/file.