mysql-5.6 VS edgelessdb

RocksDB is a fork of Google's LevelDB, a KV store using Log Structured Merge Tree (LSM tree) which is great for high write workloads. The MySQL storage engine that uses RocksDB is called MyRocks[0].

[0] http://myrocks.io/

This is much lower-level than sqlite. In fact, you could use this as the storage layer for a SQL database. See, e.g., MyRocks[0] which is a MySQL backend that uses RocksDB as the storage layer.

[0] http://myrocks.io/

> given how much $FB has benefited from running the entirety of the worlds largest social graph on top of MySQL.

They've contributed quite a lot back to the MySQL community, I thought? MyRocks comes to mind

http://myrocks.io/

as we are all discussing the pros and cons of various DBs, my favourite MySQL feature has to be the pluggable storage engines. Being able to support both InnoDB and RocksDB backed tables in the same database is wonderful. Just today I made use of the MyRocks storage engine's handy TTL [1] feature instead of a tedious delete & optimise scheduled job

[1]https://github.com/facebook/mysql-5.6/wiki/Time-to-Live-%28T...

Note: while it is based on MariaDB, it replaces InnoDB (the MariaDB/MySQL storage engine) with MyRocks (which is based on RocksDB), and as a consequence it is missing some features (such as foreign keys[1]) that prevent it from being usable in many applications.

[1] https://github.com/facebook/mysql-5.6/wiki/MyRocks-limitatio...

There's some existing work by https://www.edgeless.systems/products/edgelessdb/ and their repo https://github.com/edgelesssys/edgelessdb

The title is a bit misleading: EdgelessDB was not developed by Microsoft but by a German startup (Edgeless Systems, https://www.edgeless.systems/ and https://github.com/edgelesssys/edgelessdb).

The announcement is about its availability on the Azure Marketplace.

Hi everyone! Our team just released EdgelessDB, an open-source database built on MariaDB that runs completely inside Intel SGX enclaves. As storage engine, it uses RocksDB with a custom encryption engine. The engine uses AES-GCM and is optimized for RocksDB’s specific SST file layout and the enclave environment. It has some nice properties like global confidentiality and verifiability and it considers strong attackers like malicious admins or rootkits. It also delivers rather low overheads (<10% for the TPC-C benchmark on Azure). In short: all data is only ever decrypted inside the enclave. This is different from other databases, where data and corresponding keys are processed in the clear in memory. We believe this is useful because (1) it’s very secure and (2) it enables some interesting use cases, like secure data pooling between parties. If you’re interested in trying it out: here’s a quickstart guide. In essence, you can run the Docker image with a single command on any recent Intel Xeon with SGX. Code and more info can be found on GitHub. Would be great to get your feedback on this :-)