Is there a plan to support multi-threaded log compaction?

[MH190909]

We know that FASTER is quite fast when insert and it support single-thread log compaction.

In our use cases, the user continuously writes a large amount of data randomly, which causes the effective data to be randomly distributed across all log files. Although we can initiate single-threaded log compaction, the log compaction speed is far behind the rate of foreground user writes. This results in the system accumulating a large number of log files, meaning that the space occupied by log files is much larger than the size of the effective data set.

To address this issue and avoid excessive space amplification, we may need to consider implementing multi-threaded log recycling.

The question here is whether multi-threaded log compaction is theoretically feasible. Do we have any plans to support a multi-threaded log compaction ?

Appreciate your help!