| 3a00f978908e230f8bad034a21f5931253c10102 |
|
24-Aug-2017 |
Keun-young Park <keunyoung@google.com> |
improve I/O analysys script further
- add IO wait histogram for check_io_trace_all
- Check block level acces for check_file_read. Even if it is a single
block read in ext4 level, block level can be more efficient
or inefficient depending on I/O scheduler. Total read is coming from
block layer while ext4 layer can have more access when another read request
is made while similar read is still pending.
- add separate accounting for write
bug: 64810855
Test: run scripts and check results
Change-Id: I302d6e8661eec689bc2adb5ab608319b48acefd1
/packages/services/Car/tools/io_analysis/check_file_read.py
|
| adc951ed94e134a4c2be1b0cf659fc5b4ffc8053 |
|
17-Aug-2017 |
Keun-young Park <keunyoung@google.com> |
improve io analysis log
- add fresh read list as histogram alone was confusing / misleading due to re-read which will
mostly hit cache.
bug: 64810855
Test: capture boot time IO trace and run the script
Change-Id: I616642159a6b71cfe431b589e52576f17b5cbc08
/packages/services/Car/tools/io_analysis/check_file_read.py
|
| 96fea8b693e89892a1a90e217d8eb76bdd2edc77 |
|
01-Nov-2016 |
Keun-young Park <keunyoung@google.com> |
add scripts to analyze kernel I/O trace
- check_verity: requires custom trace in dm-verity
- check_io_trace, check_io_trace_all: requires block trace
- check_file_read: requires ext4 trace with custom trace in do_sys_open
TODO: add kernel patches for custom trace
bug: 31344213
Test: manual trace capture and analysys
Change-Id: Iae57fc017cee6fb4ccc3ea0810eac8a02ea28dbd
/packages/services/Car/tools/io_analysis/check_file_read.py
|