We use a local disk to temporarily stage data before putting it on S3. We have smaller WAL (write ahead log) objects, and a periodic compaction process which cr

We use a local disk to temporarily stage data before putting it on S3. We have smaller WAL (write ahead log) objects, and a periodic compaction process which creates read-optimized files on S3.