3 ms·
Multi-tenancy at scale with RDBMS is just hard to do. Once you get to certain volume, you just have to have different data stores for different needs. Only a su
by prabhatjha 10y ago
Multi-tenancy at scale with RDBMS is just hard to do. Once you get to certain volume, you just have to have different data stores for different needs. Only a subset of business require ACID properties so you would use RDBMS for those and for the rest choose a NoSQL stack that has the least resistance to ramp up WRT team skill set.
You can get away without using NoSQL but you probably would end up spending time and money maintaining several DB clusters, migrations, handling replication lags etc.
- dragonwriter 10y ago> Only a subset of business require ACID properties so you would use RDBMS for those and for the rest choose a NoSQL stack While ACID is an important and nice set of properties of RDBMS's, its not the only reason to choose them; its a lot more adaptable to changing query patterns and needs than less-structured document-oriented datastores. There are sometimes performance and other reasons to choose some non-relational datastore for a particular load, but "don't need ACID guarantees" is, itself, an insufficient reason for abandoning the RDBMS model for NoSQL (don't need ACID guarantees, of course, is a necessary condition to selecting a non-ACID store, but not sufficient to choose one.)
- prabhatjha 10y agoACID was just an example. Thanks for clarifying it though.