WarpStream’s Post

Shared storage architectures, on the other hand, have a lower theoretical scale ceiling, but in practice they are often much easier to scale than their shared-nothing counterparts. The reason for this is simple, but not obvious: shared storage systems separate data from metadata which introduces a layer of abstraction between the user-facing domain model and the physical sharding used by the storage engine. As a result, it is possible to choose at runtime how much of the resources we allocate to storing or retrieving data for a particular key, rather than forcing us to choose it when we create the cluster topology ➡️ https://lnkd.in/ejeF8uMt

  • No alternative text description for this image

To view or add a comment, sign in

Explore topics