A key part of building a modern data architecture is choosing "the right tool for the right job", often this means you have multiple database technologies for the different data characteristics and access patterns you have within your application. From an analytical perspective, this means not trying to optimize a single relational database to serve both transactional and analytical workloads, but rather incorporating a dedicated analytical database (this could be per tenant or 'pooled' between tenants - the partitioning model doesn't have to match that of the source database). Historically this meant creating data pipelines to keep your relational and analytical data stores in sync, but now with Zero ETL (aws.amazon.com/what-is/zero-etl/ ) you can facilitate point-to-point integration between the two (for example Amazon Aurora and Amazon Redshift). It's a great topic though and one we'll cover in a future video!
how does the saas data architecture differ when dealing with OLTP vs analytical workloads ?
A key part of building a modern data architecture is choosing "the right tool for the right job", often this means you have multiple database technologies for the different data characteristics and access patterns you have within your application. From an analytical perspective, this means not trying to optimize a single relational database to serve both transactional and analytical workloads, but rather incorporating a dedicated analytical database (this could be per tenant or 'pooled' between tenants - the partitioning model doesn't have to match that of the source database). Historically this meant creating data pipelines to keep your relational and analytical data stores in sync, but now with Zero ETL (aws.amazon.com/what-is/zero-etl/ ) you can facilitate point-to-point integration between the two (for example Amazon Aurora and Amazon Redshift). It's a great topic though and one we'll cover in a future video!