sectionRocksDB [0601014d]
- Dong, Siying, Andrew Kryczka, Yanqin Jin, and Michael Stumm. “RocksDB: Evolution of Development Priorities in a Key-Value Store Serving Large-Scale Applications.” ACM Transactions on Storage 17, no. 4 (2021): 1–32. doi.org/10.1145/3483840.
1. MyRocks
1. MyRocks
- www.youtube.com/watch?v=Lxbb_7q8iRo
- Better SSD efficiency
- Better space efficiency
-
"if you fit the entire working set in ram, you have too much ram?"
- If all working set is in RAM, then you're not being efficient with your ram
- What if we are writing back pages with one bit change? Terrible write amplification, only rows in LSM trees
- Are you trying to read keys that don't exist? B-trees have this issue
-
LSM tree is more space efficient than b-trees
- wasting 1/3rd to 1/2 of the pool, but 10% with LSM tree. How?
- Better write efficiency is that you use less space so less write efficiency
- Secondary indices?
-
RocksDB is very hard to tune
rocksdb_block_cache_sizerocksdb_max_background_compactions
- Uncommitted changes are buffered in memory by myrocks for tx's
- Performance
