Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Block device performance benching
Fetching latest commit…
Cannot retrieve the latest commit at this time.
|Failed to load latest commit information.|
Superalign ========== A tool to measure read/write performance for block devices. It performs direct (unaffected by cache or filesystem) block transfers (reads or writes) linearly or "randomly" (LFSR) while collecting timing statistics on the transfers. You get min time/max time/avg time and standard deviation, so you can perform better analysis of the received data. It has tunable parameters such as aligning a transfer on a particular boundary, adding an offset to that alignment, or doing a discard on the device prior to transfers. The tool originally started out as an addition to Arnd Bergmann's excellent flashbench tool - git://git.linaro.org/people/arnd/flashbench.git It's a bit rough, and hopefully I'll have the time to clean up the code a bit soon. Examples ======== Test 512-byte writes aligned to 4k. sa -s 1s -a 4096 /dev/mmcblk0p15 Test 1024-byte writes straddling 8k boundary. sa -s 1k -a 8192 -o 7680 /dev/mmcblk0p15 Test 8k reads. sa -s 8k -d /dev/mmcblk0p15 Test maximum 1000 different 8k writes (as opposite to as many as can fit in the device) sa -s 8192 -c 1000 /dev/mmcblk0p15 Repeat above test five times, blk erasing the device before each repeat. sa -s 8192 -c 1000 -r 5 -e /dev/mmcblk0p15 Use LFSR ("random") instead of sequential access. sa -s 8192 -R /dev/mmcblk0p15 Print verbose info. sa -v -s 8192 /dev/mmcblk0p15 Print really verbose info. sa -v -v -s 8192 /dev/mmcblk0p15 Contact Info ============ Andrei Warkentin (firstname.lastname@example.org, email@example.com)