SAGE: A logical and physical design tool for entity-group based new SQL systems

Bin Liu, Wang-Pin Hsiung, Jun'ichi Tatemura, Hakan Hacigümüs

Abstract: Entity-group based new SQL systems achieve scalability and consistency at the same time by using a key-value store as the storage layer and limiting each transaction's boundary to a collection of data (called an entity-group). Examples of such systems are Google's Megastore, NEC's Partiqle, and LinkedIn's Espresso. Application developers of such systems face tremendous challenges, both in designing entity-groups (and hence transaction boundaries) and physical layout of data in key-value stores. Entity-group designs directly impact consistency semantics of the workload, and physical layout impacts the application throughput. Both problems are challenging for users to solve manually. In this demonstration, we show a system that solves both problems with a user-friendly GUI built on our principled formal methods. We demonstrate the system using a simple Auction benchmark, and a more complex TPC-W benchmark.