Workload management of cooperatively federated computing clusters

Abstract

Cooperative resource sharing enables distinct organizations to form a federation of computing resources. The motivation behind cooperation is that organizations are likely to serve each other by trading unused CPU cycles given the existence of irregular usage patterns of their local resources. In this way, resource sharing would enable organizations to purchase resources at a feasible level while meeting peak computational throughput requirements. This federation results in community grid that must be managed. A functional broker is deployed to facilitate remote resource access within the community grid. A major issue is the problem of correlations in job arrivals caused by seasonal usage and/or coincident resource usage demand patterns. These correlations incur high levels of burstiness in job arrivals causing the job queue of the broker to grow to an extent such that its performance becomes severely impaired. Since job arrivals cannot be controlled, management strategies must be employed to admit jobs in a manner that can sustain a fair level of resource allocation performance at all participating organizations in the community. In this paper, we present a theoretical analysis of the problem of job traffic burstiness on resource allocation performance in order to elicit the general job management strategies to be employed. Based on the analysis, we define and justify a job management strategies for the resource broker to cope with overload conditions caused by job arrival correlations.

Design and implementation of an adaptive job allocation strategy for heterogeneous multi‐cluster computing systems

“…Dynamic scheduling, conducted at run time, can support dynamic load-balancing and fault-tolerance. Although load balancing shares loads among nodes in job execution [34,35], the practice introduces additional program execution overhead.…”

“…Some works consider memory latency the main indicator for load sharing [36], or take memory and CPU power usage into account [37]. Others take additional measures such as queueing [35] and coallocation method [38] into consideration. Previously, there was no need to consider co-allocationallocation of jobs across clusters-in a single-cluster system, but it must be implemented in multi-cluster environments.…”

An Adaptive Job Allocation Strategy for Heterogeneous Multiple Clusters

“…As to the dynamic scheduling, it conducts the scheduling at run time and it can support dynamic load balancing and fault tolerance. Although load balancing could share the load between nodes after the job is in its execution state [26] [27].…”

“…Some works considers the memory latency as the main indicator for load sharing [29], or taking both memory and CPU power usage into account [27][28] [30]. Previously, on a single cluster system, there is no need of considering coallocation which is allocating jobs across different clusters.…”

On construction of a well-balanced allocation strategy for heterogeneous multi-cluster computing environments

“…Some of them use memory latency as the main indicator for load sharing [17], or take memory and CPU power usage into account [6]. Others take further steps like queuing [21] or co-allocation [11] into consideration. In single-cluster systems there is no need to consider coallocation, which allocates jobs across clusters, but it is necessary in multi-cluster systems.…”