Randomized Load Balancing in Finite Regimes

Godtschalk, Antonie S.; Ciucu, Florin

doi:10.1109/icdcs.2016.97

Cited by 2 publications

(3 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Authors provide upper bounds of the discard probability for when N, C, λ → ∞. Another study [29] used SQ(d) for a finite set N of M/M/1 queues. They provide numerical bounds on the delay and show that the delay can be quite different w.r.t the limit case.…”

Section: Background and Related Workmentioning

confidence: 99%

A Power-of-Two Choices Based Algorithm for Fog Computing

Beraldi

Alnuweiri

Mtibaa

2020

IEEE Trans. Cloud Comput.

View full text Add to dashboard Cite

The fog computing paradigm brings together storage, communication, and computation resources closer to users' end-devices. Therefore, fog servers are deployed at the edge of the network, offering low latency access to users. With the expansion of such fog computing services, different providers will be able to deploy multiple resources within a restricted geographical proximity.In this paper, we investigate an incentive-based cooperation scheme across fog providers. We propose a distributed cooperative algorithm amongst fog computing providers where fully collaborative fog nodes are subject to different loads. The proposed algorithm leverages the power-of-two result and exploits a cooperation probability, namely the probability that a given provider collaborates by accepting a computation request from another provider, as a mean to achieve a fair cooperation.We adopt an analytical approach based on exploiting a simplified performance model to demonstrate numerically that a set of optimal accepting probabilities exits when the number of server nodes goes to infinity. This result then drives the design of our distributed algorithm. Second, in our experimental approach, we perform a set of simulation analysis to verify the validity of the proposed solution when the number of servers is limited.

show abstract

Section: Background and Related Workmentioning

confidence: 99%

A Power-of-Two Choices Based Algorithm for Fog Computing

Beraldi

Alnuweiri

Mtibaa

2020

IEEE Trans. Cloud Comput.

View full text Add to dashboard Cite

show abstract

“…A wide range of randomized strategies including the schemes 1 and 2 presented in [5] can be viewed as CBS strategies. The first prominent randomized strategy was the power of two strategy [4], [13], which was also generalized to the power of d strategy or the so-called SQ(d) strategy [14], [15]. A survey of SQ(d)-type load balancing strategies for large systems is provided in [16].…”

Section: Related Workmentioning

confidence: 99%

“…This asymptotic estimate is remarkable, but it is not necessarily accurate for a finite number of servers. In this case, upper and lower bounds on the average delay are provided in [15]. Additionally, in [17] a bound for the queue lengths is estimated for the supermarket model under appropriate arrival rate scaling.…”

Section: Related Workmentioning

confidence: 99%

Generalized Cost-Based Job Scheduling in Very Large Heterogeneous Cluster Systems

KhudaBukhsh

Kar

Alt

et al. 2020

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

We study job assignment in large, heterogeneous resource-sharing clusters of servers with finite buffers. This load balancing problem arises naturally in today's communication and big data systems, such as Amazon Web Services, Network Service Function Chains, and Stream Processing. Arriving jobs are dispatched to a server, following a load balancing policy that optimizes a performance criterion such as job completion time. Our contribution is a randomized Cost-Based Scheduling (CBS) policy in which the job assignment is driven by general cost functions of the server queue lengths. Beyond existing schemes, such as the Join the Shortest Queue (JSQ), the power of d or the SQ(d) and the capacity-weighted JSQ, the notion of CBS yields new application-specific policies such as hybrid locally uniform JSQ. As today's data center clusters have thousands of servers, exact analysis of CBS policies is tedious. In this work, we derive a scaling limit when the number of servers grows large, facilitating a comparison of various CBS policies with respect to their transient as well as steady state behavior. A byproduct of our derivations is the relationship between the queue filling proportions and the server buffer sizes, which cannot be obtained from infinite buffer models. Finally, we provide extensive numerical evaluations and discuss several applications including multi-stage systems.

show abstract

Randomized Load Balancing in Finite Regimes

Cited by 2 publications

References 21 publications

A Power-of-Two Choices Based Algorithm for Fog Computing

A Power-of-Two Choices Based Algorithm for Fog Computing

Generalized Cost-Based Job Scheduling in Very Large Heterogeneous Cluster Systems

Contact Info

Product

Resources

About