Unmanaged Multiversion STM ∗

Multiversion STM systems allow a transaction to read old values of a recently updated object, after which the transaction may serialize before transactions that committed earlier in physical time. This ability to “commit in the past” is particularly appealing for long-running read-only transactions, which may otherwise starve in many STM systems, because short-running peers modify data out from under them before they have a chance to finish. Most previous work on multiversion STM assumed an object-oriented, garbage-collected language. In contrast, we present a multiversion STM system designed for unmanaged languages, in which ownership information is typically maintained on a “per-slice” (rather than a “per-object”) basis, and in which automatic garbage collection may not be available. We have implemented this UMV system as an extension to the TL2-like LLT back end of the RSTM suite. Experiments with microbenchmarks confirm that UMV can dramatically reduce the abort rate of long read-only transactions. Overall impacts range from 2× slowdown to 5× speedup, depending on the offered workload.