I don't know the actual reason internal to Pinterest, but if I have to guess... It seems like they are using this DB for analytical use-cases. In these scenario, the database is written to much more often than its read from. LSM Tree indexes are great for writing at scale in comparison to B-Tree indexes (in Relational databases). The strong relational data consistency guarantees are also not required here since it should be okay if there are few inconsistent data-points in a sea of data. (I have seen postgres being for a scenario like this and the data-model becomes quite complex. Although performance was decent, I am not sure how it would fare at Pinterest's scale) Of course, there might be multitude of other reasons, like ease of scaling, simpler data model, team expertise, established internal frameworks and infra etc.
![[กูเพิ่งซื้อมึงปิดหนี] 20 นาที ผมรีวิวเกมที่แย่ที่สุดในโลก CONCORD](http://i.ytimg.com/vi/XhC-3q3Ii7I/mqdefault.jpg)
great explanation. very helpful. thanks a lot for this.
Thanks for watching!
Great thank you!
Glad it was helpful!
Hey really great video 🙌
Thanks for the visit
I don't get why they don't use a relational database. Speed?
I don't know the actual reason internal to Pinterest, but if I have to guess... It seems like they are using this DB for analytical use-cases. In these scenario, the database is written to much more often than its read from. LSM Tree indexes are great for writing at scale in comparison to B-Tree indexes (in Relational databases). The strong relational data consistency guarantees are also not required here since it should be okay if there are few inconsistent data-points in a sea of data.
(I have seen postgres being for a scenario like this and the data-model becomes quite complex. Although performance was decent, I am not sure how it would fare at Pinterest's scale)
Of course, there might be multitude of other reasons, like ease of scaling, simpler data model, team expertise, established internal frameworks and infra etc.
Great explanation
1 month no upload 🫠
New video on geosharding and hilbert curve coming soon! Was dealing with some stuff lately