Transactional scheduling for read-dominated workloads

Abstract

International audienceThe transactional approach to contention management guarantees atomicity by aborting transactions that may violate consistency. A major challenge in this approach is to schedule transactions in a manner that reduces the total time to perform all transactions (the makespan), since transactions are often aborted and restarted. The performance of a transactional scheduler can be evaluated by the ratio between its makespan and the makespan of an optimal, clairvoyant scheduler that knows the list of resource accesses that will be performed by each transaction, as well as its release time and duration. This paper studies transactional scheduling in the context of read-dominated workloads; these common workloads include read-only transactions, i.e., those that only observe data, and late-write transactions, i.e., those that update only towards the end of the transaction. We present the Bimodal transactional scheduler, which is especially tailored to accommodate read-only transactions, without punishing transactions that write most of their duration (early-write transactions). It is evaluated by comparison with an optimal clairvoyant scheduler; we prove that Bimodal demonstrates the best competitive ratio achievable by a non-clairvoyant schedule for workloads consisting of early-write and read-only transactions. We also show that late-write transactions significantly deteriorate the competitive ratio of any non- clairvoyant scheduler, assuming it takes a conservative approach to conflicts

Window-based greedy contention management for transactional memory: theory and practice

“…The makespan of the transactional scheduling algorithm can be compared to the makespan of an optimal off-line scheduling algorithm (which has the complete knowledge of the resource requests along with the arrival and execution time durations of transactions) to provide a competitive ratio. The makespan and the competitive ratio primarily depend on the workload-the set of transactions, along with their arrival times, execution time durations, and resources they read and modify [4].…”

“…Several other work studied contention managers for the efficient contention management in STM systems [2,3,11,12,[16][17][18]24,37,40,46] and the performance of some of them has been analyzed formally in [3,4,12,16,18,24,40]. These works applied several different models and techniques for analyzing transactional memory performance.…”

“…On the other side, TM schedulers [2,4,11,12,46] offer an alternative approach to boost the TM performance. TM scheduler is, basically, a software component which decides when a particular transaction executes.…”

“…Of particular interest are greedy contention managers where a transaction starts again immediately after every abort. Several (greedy and non-greedy) contention managers have been proposed in the literature [2][3][4]11,12,16,18,24,[37][38][39][40]46].…”

“…Empirical studies on different contention manager proposals ranging from simple exponential back-off to complex priority-based techniques have shown that the choice of a contention manager can significantly affect the performance of the STM systems, sometimes drastically [18,26,27,38,39]. There is a small amount of work in the literature which analyzes formally the theoretical performance of contention managers [3,4,12,16,18,24,40] (details in Sect. 2).…”