Interference-Aware Scheduling Using Geometric Constraints

Abstract

The large scale parallel and distributed platforms produce a continuously increasing amount of data which have to be stored, exchanged and used by various jobs allocated on different nodes of the platform. The management of this huge communication demand is crucial for the performance of the system. Meanwhile, we have to deal with more interferences as the trend is to use a single all-purpose interconnection network. In this paper, we consider two different types of communications: the flows induced by data exchanges during computations and the flows related to Input/Output operations. We propose a general model for interference-aware scheduling, where explicit communications are replaced by external topological constraints. Specifically, we limit the interferences of both communication types by adding geometric constraints on the allocation of jobs into machines. The proposed constraints reduce implicitly the data movements by restricting the set of possible allocations for each job. We present this methodology on the case study of simple network topologies, namely the line and the ring. We propose theoretical lower and upper bounds under different assumptions with respect to the platform and jobs characteristics. The obtained results illustrate well the difficulty of the problem even on simple topologies.

A Methodology for Handling Data Movements by Anticipation: Position Paper

“…Our proposal was to identify relevant constraints that can easily be integrated into an optimization problem. We succesfuly applied this methodology for a specic topology (line/ring of processors) [7]. We are currently working on the design of a generic heuristic that can address several topologies.…”

Scheduling Malleable Jobs Under Topological Constraints

“…If this machine is allocated to a different job, then we have the undesirable effect of delaying the completion time of one job in order to handle the traffic from a different job. In order to avoid both the aforementioned data flows, we use the following definitions introduced in [3], [4] to restrict the number of possible allocations of a job. Definition 1.…”

“…Note that, Bleuse et al [3] presented a 6-approximation algorithm for the problem of scheduling rigid jobs under the contiguity and the locality constraints. However, these constraints have not been studied in the context of scheduling malleable jobs, which is the main subject of our paper.…”

“…However, if m is part of the instance, there is no polynomial time approximation algorithm with ratio smaller that n δ , for any δ > 0, as shown in [11]. Although this negative result for the more general P |set j |C max , a 6-approximation algorithm has been presented in [3] for scheduling rigid jobs under contiguity and locality constraints.…”

“…Extending this work, Lucarelli et al studied the impact in backfilling scheduling of topological constraints like contiguity and locality in hierarchical platforms [2]. Bleuse et al [3], [4] introduced a more general model for interference-aware scheduling in large scale parallel platforms by distinguishing between direct and indirect topologies. Moreover, they study the effect of a second type of nodes, the input/output nodes, in conjunction with standard computing nodes.…”