You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have confirmed that this should be classified as an enhancement rather than a bug/feature.
Summary
We will use rocksdb's time wheel support to implement arbitrary delayed messages, solve the problem of cold read and metric imprecision caused by large-scale delayed messages, and support timely deletion and more accurate latency levels and other features
Motivation
We have encountered some problems in the actual production environment, and have been troubled by these problems. In large-scale timed message scenarios, frequent reading of old files reduces the read rate, increases the delay of delayed messages, and causes problems such as inaccurate observation and delay. Meanwhile, functions such as preheating of file version suffer large losses. Therefore, the introduction of rocksdb can avoid the impact of cold read, provide more accurate index calculation, and support longer timed messages. Provide more precise latency levels, but we still need to ensure message write performance and read performance and latency, etc.
Before Creating the Enhancement Request
Summary
We will use rocksdb's time wheel support to implement arbitrary delayed messages, solve the problem of cold read and metric imprecision caused by large-scale delayed messages, and support timely deletion and more accurate latency levels and other features
Motivation
We have encountered some problems in the actual production environment, and have been troubled by these problems. In large-scale timed message scenarios, frequent reading of old files reduces the read rate, increases the delay of delayed messages, and causes problems such as inaccurate observation and delay. Meanwhile, functions such as preheating of file version suffer large losses. Therefore, the introduction of rocksdb can avoid the impact of cold read, provide more accurate index calculation, and support longer timed messages. Provide more precise latency levels, but we still need to ensure message write performance and read performance and latency, etc.
Detailed design: https://shimo.im/docs/913JMO8ZoJugwEAE
Describe the Solution You'd Like
The proposal addresses two main issues:
Extra can support more optimizations
Describe Alternatives You've Considered
no
Additional Context
No response
The text was updated successfully, but these errors were encountered: