RethinkDB seems to implement a functional btree on a log-structured storage. This would involve lots of copying, I wonder what is the performance impact of this (compared to a less pure implementation that copies the modified nodes every thousand updates or so).
Also, the most important feature for SSD storage (no random writes!) is fulfilled by Cassandra / BigTable implementations based on SSTables, I wonder how a btree-based implementation compares to them.
Also, the most important feature for SSD storage (no random writes!) is fulfilled by Cassandra / BigTable implementations based on SSTables, I wonder how a btree-based implementation compares to them.
reply