Unlike SQL designs that build on K/V stores, this allows to DSQL to do much of the heavy lifting of filtering and finding data right next to the data itself, on the storage replicas, without sacrificing scalability of storage or compute.