Tigon: A Distributed Transactional Database for a CXL Pod

Abstract: Building an efficient distributed transactional database remains a challenging problem despite decades of research. Existing distributed databases synchronize cross-host data accesses via networks, which requires numerous network message exchanges and introduces inevitable overhead. In this talk, I will describe Tigon, the first distributed transactional database that synchronizes cross-host concurrent data accesses via atomic operations on CXL memory. I will discuss the limitations of CXL memory: its higher latency and lower bandwidth relative to local DRAM, and its limited support for hardware cache coherence across hosts. and then present how Tigon addresses these limitations. Using TPC-C and YCSB benchmarks, Tigon achieves up to 2.8x higher throughput compared to an optimized shared-nothing database that uses CXL memory as transport and up to 14.4x higher throughput compared to an RDMA-based distributed database. The paper for Tigon will appear in OSDI 2025.

Bio: Dixin Tang is an Assistant Professor at the CS department of UT Austin. Before joining UT Austin, he worked as a postdoc scholar at UC Berkeley and earned his Ph.D. from the University of Chicago. His research interests lie in building efficient and scalable data systems for modern hardware and applications. 

-- For the zoom passcode, contact the organizer at markakis@mit.edu