3 ms·
The nugget of truth was that joins and transactions are very difficult to implement in a performant manner in large scale, distributed systems. So it was possi
by jimbokun 7y ago
The nugget of truth was that joins and transactions are very difficult to implement in a performant manner in large scale, distributed systems. So it was possible to write very highly scalable databases if you give up on "SQL", which was a stand in for joins and transactions.
Since then, developers have been coming up with clever ways to implement distributed joins and transactions more efficiently and scalably on top of distributed database engines, and thus "the return of SQL".
- 6gvONxR4sf7o 7y agoWhere can I learn more about these clever developments?
- jimbokun 7y agoI find CockroachDB one of the most interesting systems adding distributed transactions, joins and SQL support on top of a highly scalable key value store engine: https://www.cockroachlabs.com/docs/stable/architecture/overview.html https://www.cockroachlabs.com/docs/stable/architecture/overv...
- jwatte 7y agoFor example: http://info.snowflake.net/rs/252-RFO-227/images/Snowflake_SIGMOD.pdf http://info.snowflake.net/rs/252-RFO-227/images/Snowflake_SI... https://www.snowflake.com/wp-content/uploads/2014/10/A-Detailed-View-Inside-Snowflake.pdf https://www.snowflake.com/wp-content/uploads/2014/10/A-Detai...