feat(rocksdb): Adapt rate limiter to prevent bust writes and huge compaction

[foreverneverer]

What problem does this PR solve?

rocksdb compaction has large influence when put data. this pr add rocksdb write limiter to improve performance.The follows are some result:

1. no limit, 3(client)*20(thread)

2. 500MB limit and open auto-tune, 3(client)*20(thread)

3. 500MB limit and open auto-tune, 3(client)*30(thread)

Besides, for observe the pegasus performance, this pr add one new perfcounter: rate limiter through rate: replica*app.pegasus*rdb.write_limiter_rate_bytes

And add new config:

[pegasus.server]
rocksdb_limiter_max_write_megabytes_per_sec = 500
rocksdb_limiter_enable_auto_tune = false

Note: rocksdb_limiter_enable_auto_tune = true may cause write stall especially when the put data size is large(In my test, it will happen when qps=10k, value=10k), so new commit set the config to false

What is changed and how it works?

add rocksdb write limiter for compaction and flush

Check List

Tests

• Manual test (add detailed scripts or steps below)

Related changes

• Need to cherry-pick to the release branch
• Need to be included in the release note
• Need to update config

[acelyc111]

How about rename this pr to "feat(rocksdb): Adapt rate limiter to prevent bust writes and huge compaction"?

[foreverneverer]

Note: rocksdb_limiter_auto_tune_enable = true may cause write stall especially when the put data size is large(In my test, it will happen when qps=10k, value=10k), so new commit set the config to false

Update the description

[acelyc111]

How about rename this pr to "feat(rocksdb): Adapt rate limiter to prevent bust writes and huge compaction"?

@shuo-jia

[foreverneverer]

How about rename this pr to "feat(rocksdb): Adapt rate limiter to prevent bust writes and huge compaction"?

@shuo-jia

Yeah，have updated