[4pt] Investigate merger/key_def embedding into current implementation about crud HOT 4 CLOSED

I would be interested in performance testing results.

Here is my modest benchmarks - #38 (comment)

from crud.

See https://github.com/Totktonada/tarantool-merger-examples/tree/master/chunked_example_fast

from crud.

I would be interested in performance testing results.

A measurement may be more close to a real life if we would imitate a delay between a router and a storage:

tc qdisc add dev lo root handle 1:0 netem delay 500msec  # 1ms rtt
tc qdisc del dev lo root  # clean up

However microbenchmark is interesting too.

from crud.

I have simple benchmark - https://github.com/tarantool/crud/tree/perf-test

But at first glance I see some slowdown after merger integration (introduced here #38).
Seems we need to cache key_def instances (Currently this optimization is switched off). Currently we have implementation that passes our tests but further investigation is required to estimate and solve performance problems.

from crud.