Tianzheng Wang Ryan Johnson Alan Fekete Ippokratis

Transcript:Tianzheng Wang Ryan Johnson Alan Fekete Ippokratis:
Tianzheng Wang Ryan Johnson Alan Fekete Ippokratis Pandis The Serial Safety Net: Efficient concurrency control on modern hardware 1 Modern HW: CPU on critical path 2 Low parallelism Focus: hide I/O stalls Classic DBMS Modern DBMS High parallelism Focus: minimize cycles Main-memory OLTP: more pressure on CC Locks Disks Threads Memory Disk Threads CC: current state of the art OCC (fast) SI (non-serializable) SSI (slow) OCC looks good. What’s the problem? Contentious TPC-C variant 24 16 6 12 1 3 CC: Robustness matters! 4 OCC (unfair) SI (non-serializable) SSI (slow) Need all three: fast, fair, serializable 24 16 6 12 1 Contentious TPC-C variant with retries The Serial Safety Net (SSN) 5 Some CC >= RC SSN Serializable execution Might admit anomalies Cheap certifier Works even if CC is buggy Aborts offenders Dictates access pattern SSN: fast, fair, serializable 6 OCC SI SSI SSN Database model 7 R1 R2 R3 R4 ... R3 Uncommitted write Invisible to other readers Multi-version database R1 R1 CC decides which version to read Captures most CC schemes, incl. 2PL Serialization graphs and cycles T1 r:y:w T2 w:x:r T3 r:x:w T1 8 SI read-only anomaly T3 T2 T1 (T3) commit time dependency order Serialization graphs and cycles T1 commits last & “closes” the cycle “Hard” to detect T1 … T3 “Easy” to detect T3  T1 SSN: efficiently detect T1 … T3 9 SSN in a nutshell 10 Define π(T) = min {c(S) : T … S} Define c(T) = “commit time of T” Forbid $P  T : π(T) ≤ c(P) ≤ c(T) Track T’s earliest successor P might be a successor of T T might close a cycle Visualizing SSN 11 Exclusion window satisfied commit time dependency order Evaluation System 4 x 6-core Xeon @ 1.8GHz, 64GB DRAM A variant of Silo* that supports SI and RC TPC-C w/ random warehouse selection RC, SI, OCC (Silo), SSI, {SI, RC} + SSN What to test General performance/scalability Fairness for updates vs. reads Robustness under retries 12 * Tu et al, “Speedy transactions in multicore in-memory databases”, SOSP`13 SSN performs well 13 Higher is better SSI OCC Low implementation overhead SSN has low abort rate 14 Lower is better OCC SI (higher) and RC/SI+SSN SSI SSN allows more valid schedules than SSI or OCC SSN provides “safe retry” 15 OCC SI (highest) and RC/SI+SSN SSI Higher is better SSN is fair 16 SSI starves writers