Back to Search Start Over

Optimal hyper-scalable load balancing with a strict queue limit

Authors :
Sem Borst
Mark van der Boor
Johan S. H. van Leeuwaarden
Stochastic Operations Research
Econometrics and Operations Research
Research Group: Operations Research
Source :
Performance Evaluation, 149-150:102217. Elsevier
Publication Year :
2021

Abstract

Load balancing plays a critical role in efficiently dispatching jobs in parallel-server systems such as cloud networks and data centers. A fundamental challenge in the design of load balancing algorithms is to achieve an optimal trade-off between delay performance and implementation overhead (e.g. communication or memory usage). This trade-off has primarily been studied so far from the angle of the amount of overhead required to achieve asymptotically optimal performance, particularly vanishing delay in large-scale systems. In contrast, in the present paper, we focus on an arbitrarily sparse communication budget, possibly well below the minimum requirement for vanishing delay, referred to as the hyper-scalable operating region. Furthermore, jobs may only be admitted when a specific limit on the queue position of the job can be guaranteed. The centerpiece of our analysis is a universal upper bound for the achievable throughput of any dispatcher-driven algorithm for a given communication budget and queue limit. We also propose a specific hyper-scalable scheme which can operate at any given message rate and enforce any given queue limit, while allowing the server states to be captured via a closed product-form network, in which servers act as customers traversing various nodes. The product-form distribution is leveraged to prove that the bound is tight and that the proposed hyper-scalable scheme is throughput-optimal in a many-server regime given the communication and queue limit constraints. Extensive simulation experiments are conducted to illustrate the results.

Details

Language :
English
ISSN :
01665316
Database :
OpenAIRE
Journal :
Performance Evaluation, 149-150:102217. Elsevier
Accession number :
edsair.doi.dedup.....f642fdf076cb8331b4867fa1135d18fd