Probabilistic job symbiosis modeling for SMT processor scheduling

Abstract

Symbiotic job scheduling boosts simultaneous multithreading (SMT) processor performance by co-scheduling jobs that have 'compatible' demands on the processor's shared resources. Existing approaches however require a sampling phase, evaluate a limited number of possible co-schedules, use heuristics to gauge symbiosis, are rigid in their optimization target, and do not preserve system-level priorities/shares. This paper proposes probabilistic job symbiosis modeling, which predicts whether jobs will create positive or negative symbiosis when co-scheduled without requiring the co-schedule to be evaluated. The model, which uses per-thread cycle stacks computed through a previously proposed cycle accounting architecture, is simple enough to be used in system software. Probabilistic job symbiosis modeling provides six key innovations over prior work in symbiotic job scheduling: (i) it does not require a sampling phase, (ii) it readjusts the job co-schedule continuously, (iii) it evaluates a large number of possible co-schedules at very low overhead, (iv) it is not driven by heuristics, (v) it can optimize a performance target of interest (e.g., system throughput or job turnaround time), and (vi) it preserves system-level priorities/shares. These innovations make symbiotic job scheduling both practical and effective. Our experimental evaluation, which assumes a realistic scenario in which jobs come and go, reports an average 16% (and up to 35%) reduction in job turnaround time compared to the previously proposed SOS (sample, optimize, symbios) approach for a two-thread SMT processor, and an average 19% (and up to 45%) reduction in job turnaround time for a four-thread SMT processor

Mitigating execution unit contention in parallel applications using instruction‐aware mapping

“…Feliu et al's 17 work implements a simplification of the probabilistic model proposed by Eyerman et al, 19 which is infeasible since it would require much more per‐core hardware counters than the processors can deliver. Feliu et al mitigate that by using per core CPI stacks available in IBM POWER processors.…”

“…However, for applications that stress floating-point and integer units, their mechanism can lead to non-optimal scheduling. 19 which is infeasible since it would require much more per-core hardware counters than the processors can deliver. Feliu et al mitigate that by using per core CPI stacks available in IBM POWER processors.…”

Bubble-flux

“…Because the Dynamic Bubble operates online, it can capitalize on variations in application sensitivity due to load fluctuations to increase utilization. Scheduling techniques for QoS management are proposed for both CMPs [14,17,21,38,42] and SMT architectures [13,35]. Other OS mechanisms such as page-coloring are also proposed [23].…”

Optimal task assignment in multithreaded processors

“…The SOS scheduler uses hardware performance counters to find schedules that exhibit good performance. Eyerman and Eeckhout [22] propose probabilistic job symbiosis model that enhances the SOS scheduler. Based on the cycle accounting architecture [21], the model estimates the singlethreaded progress for each job in a multithreaded workload.…”