Meeting Minutes and Working Plan (made on 2019.01.16)

Meeting Minutes

• I should give enough evidence that the observations are not likely to happen in normal situations.

• There are some performance bug caused by the mismatch between FS and DB. I will study on that.

After the meeting, I make it clear that what subjects of software system should I focus on. (that is Storage system, including databases & filesystems) Also, I know what kind of performance bug should I focus on. (that is configuration related performance bugs, in other words, performance bugs that can be revealed by configuration variation). Further, since the 4 observations(e.g. 0 workloads but non-0 cost) are all related to configuration, I focus only on the performance against variation on configuration in stead of workload&configuration(This is because workloads are too complicated). In this way, we can combine these 4 observations(surge points, 0 workloads but non-0 cost,...) into one -- Counter-intuitive(by comparing 'supposed' performance of a configuration and real performance of a configuration under some benchmark tests).

I will summarize the full idea soon

• Problems in the meeting are listed in Issues.

Working Plan

1. Elaborate methodology of my empirical study. (week 1-2)

2. Study related works of Performance bugs in FSs (read papers). (week 1-2)

3. Carry out the study on Performance bugs in FSs (e.g. in LKML). (week 3-)

4. Keep on studying Performance bugs in databases.

@shanshanL @apollozzz @jostep