4 ms·
> New problem: Your partitions need to be < 300 MB or cassandra falls over dead. people work around that by storing the data in something like S3 and then keep
by MichaelMoser123 6y ago
> New problem: Your partitions need to be < 300 MB or cassandra falls over dead.
people work around that by storing the data in something like S3 and then keeping the handle in cassandra. Yet another hack on top of another hack.
- aasasd 6y agoPretty sure that's what the article is about.
- AtlasBarfed 6y agoDo you fill postgres with 100MB+ blobs? You'd do the same "hack" there.
- MichaelMoser123 6y agoyes, but postgress doesn't have the problem that it fails if you have more than 100MB in all records, in total.
- PeterCorless 6y agoWhile part of this is Cassandra's architecture, it may also be a limitation of the designer's way of thinking. Recently we had a hackathon project where we decided to implement the s3 API on top of Scylla (a Cassandra-compatible database) and were able to chunkify binary large objects (BLOBs) of up to 5 TB, and even turn Scylla into a FUSE filesystem. https://www.scylladb.com/2020/12/15/scylladb-blog-scylladb-developer-hackathon-scylla-s3/ https://www.scylladb.com/2020/12/15/scylladb-blog-scylladb-d...