> This write amplification is large enough that our efforts to tune indexing throughput have started to hit diminishing returns.
> don't key on the ANN address. That is precisely the change turbopuffer v3 makes. As you can imagine, it is not a trivial change.
This is a direct parallel to how Postgres and Mysql built indexes.
Your design choice went from a Postgres design pattern to a Mysql one. The difference is the reindexing cost vs the lookup cost - Postgres optimized for lookup and Mysql does for indexing on writes. Or more accurately, Postgres was better with good schema design using joins & mysql was optimized for a bad design with less normalization where many indexes exist for the same table.
Postgres always points an index to a row-id within postgres which is an arbitrary value which changes on each update.
Mysql, always assuming the storage engine is pluggable, points to the primary index entry and adds an extra indirection to the lookup.
This means that you point the mysql index to a stable id, so unless you go update the primary key for a row, you won't have to update the indexes for all the attribute lookups you might have made to data.
I don't do databases any more that much, but the design for NIMBLE file format has a lot of quirks which are relevant to this specific idea (wide tables).
But the old Uber post about switching from Postgres to Mysql to prevent index amplification[1] is a direct mirror to this post.
I literally had opening slide about Worse is Better when adding mysql support to peerdb. Lots of things in mysql have v1 as stupidest thing that works, then v2 fixes issues
There's a bunch of internal types like decimal vs newdecimal, binlog started out statement based until they realized uuid generation is random so added data replication on top. CDC offset started as filepos before GTID was made so offset could survive failover
There were aspects of the design I appreciated (logical slots in postgres have a bunch of drawbacks avoided by just appending to 2nd serial log which has an expiry date instead of tracking clients' offsets), but developing against protocol you learn to not try build a consistent mental model
MySQL had worse-is-better dominance over postgres until, like, 2016 or so? I'll take the worse solution with better performance and a better replication story any day.
I hear you, but there's just so many knobs on postgres that at our correct scale we couldn't amortize the time and effort to become proficient at postges.
gopalv · · focus · HN ↗
> don't key on the ANN address. That is precisely the change turbopuffer v3 makes. As you can imagine, it is not a trivial change.
This is a direct parallel to how Postgres and Mysql built indexes.
Your design choice went from a Postgres design pattern to a Mysql one. The difference is the reindexing cost vs the lookup cost - Postgres optimized for lookup and Mysql does for indexing on writes. Or more accurately, Postgres was better with good schema design using joins & mysql was optimized for a bad design with less normalization where many indexes exist for the same table.
Postgres always points an index to a row-id within postgres which is an arbitrary value which changes on each update.
Mysql, always assuming the storage engine is pluggable, points to the primary index entry and adds an extra indirection to the lookup.
This means that you point the mysql index to a stable id, so unless you go update the primary key for a row, you won't have to update the indexes for all the attribute lookups you might have made to data.
I don't do databases any more that much, but the design for NIMBLE file format has a lot of quirks which are relevant to this specific idea (wide tables).
But the old Uber post about switching from Postgres to Mysql to prevent index amplification[1] is a direct mirror to this post.
[1] - <a href="https://www.uber.com/us/en/blog/postgres-to-mysql-migration/" rel="nofollow">https://www.uber.com/us/en/blog/postgres-to-mysql-migration/
phoghed · · focus · HN ↗
TIL I should have been using mysql the whole time
woadwarrior01 · · focus · HN ↗
__s · · focus · HN ↗
There's a bunch of internal types like decimal vs newdecimal, binlog started out statement based until they realized uuid generation is random so added data replication on top. CDC offset started as filepos before GTID was made so offset could survive failover
There were aspects of the design I appreciated (logical slots in postgres have a bunch of drawbacks avoided by just appending to 2nd serial log which has an expiry date instead of tracking clients' offsets), but developing against protocol you learn to not try build a consistent mental model
nixon_why69 · · focus · HN ↗
jgalt212 · · focus · HN ↗