[ltt-dev] [RFC PATCH] block: Fix bio merge induced high I/O latency

[bgamari@gmail.com (Ben Gamari)]

On Sat, 2009-01-17 at 11:26 -0500, Mathieu Desnoyers wrote:
> This patch implements a basic test to make sure we never merge more than 128
> requests into the same request if it is the "last_merge" request. I have not
> been able to trigger the problem again with the fix applied. It might not be in
> a perfect state : there may be better solutions to the problem, but I think it
> helps pointing out where the culprit lays.

Unfortunately, it seems like the patch hasn't really fixed much. After
porting it forward to Linus' master, I haven't exhibited any difference
in real world use cases (e.g. desktop use cases while building a
kernel).

Given Jen's remarks, I suppose this isn't too surprising. Does anyone
else with greater familiarity with the block I/O subsystem have any more
ideas about the source of the slowdown? It seems like the recent patches
incorporating blktrace support into ftrace could be helpful for further
data collection, correct?

- Ben