Self-tuning speculation for maintaining the consistency of client-cached data

This paper presents a new protocol, Self-tuning Active Data-aware Cache Consistency (SADCC), which employs parallel communication and self-tuning speculation to improve the performance of data-shipping database systems. Using parallel communication with simultaneous client-server and client-client communication, SADCC reduces the network latency for detecting data conflicts by 50%, while increasing message volume overhead by only about 4.8%. By being aware of the global states of cached data, clients self-tune between optimistic and pessimistic consistency control. The abort rate is reduced by statistically quantifying the speculation cost. We compare SADCC with the leading cache consistency algorithms, Active Data-aware Cache Consistency (ADCC) and Asynchronous Avoidance-based Cache Consistency (AACC), in a page server DBMS architecture with page-level consistency. The experiments show that, in a non-contention environment, both SADCC and ADCC display a slight reduction (an average of 2.3%) in performance compared to AACC with a high-speed network environment. With high contention, however, SADCC has an average of 14% higher throughput than AACC and 6% higher throughput than ADCC.