Simulation of job allocation in distributed processing systems

Konovalov, Mikhail; Razumchik, Rostislav

doi:10.1109/icumt.2014.7002163

Cited by 4 publications

(4 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although in some cases the values of the job's mean response time can be computed explicitly from the available analytic results, the values displayed in the tables were obtained from the simulation (the simulation framework is described in Konovalov and Razumchik (2014); Konovalov (2014)). In the implementation of the JSQ that was used, the ties were broken according to the rule: if two or more queues have the same number of jobs, choose the queue with the fastest server; if there is no such queue, choose the random one.…”

Section: Numerical Experimentsmentioning

confidence: 99%

ECMS 2019 Proceedings edited by Mauro Iacono, Francesco Palmieri, Marco Gribaudo, Massimo Ficco

2019

View full text Add to dashboard Cite

The 33rd edition of the ECMS conference, a widely consolidated interdisciplinary forum for simulation and modeling researchers and practitioners that keeps its reference position year after year, brings with it two novelties: a new name and a new form of publication. In fact, to better catch in the name the actual situation of participants, the European Council for Modelling and Simulation decided to rename the event as "ECMS International Conference on Modelling and Simulation". This year, we also start a new book series of annual volumes, namely "Communications of the ECMS", which replaces the name "Proceedings" in order to acknowledge the recognition and reputation of the scientific community of simulation. This year the chosen conference venue is in Caserta, Italy. The Università degli Studi della Campania "Luigi Vanvitelli", previously known as Seconda Università degli Studi di Napoli, a state-owned university founded in 1990 with the explicit mission of meeting the cultural demand of the population living in the northern part of Campania. Caserta, famous for its wonderful royal palace, whose facilities will host the conference, is a city with a long history, already inhabited in the 5th century BC. It flourished in the middle age and then in the XVI-XVII century AD, surrounded by a geographical area rich of historical relevance that was under Roman influence.The university is characterized by its diffusion over the territory of the province of Caserta, having its 16 departments located in Caserta, Aversa and Santa Maria Capua Vetere, and Napoli, in which its roots can be found: the most of the departments are located in historical landmarks, that have been saved and recovered by specific agreements and funding. More than 25000 students from Italy and abroad study at Vanvitelli University enrolled in bachelor, master and PhD. curricula, taught as well in Italian as in English, and covering different aspects of mathematics, physics, data analytics, engineering, law, economics, political sciences, medicine, nursery, biotechnology, pharmacy, environmental sciences, cultural heritage, literature, psychology, tourism, design and architecture.We are honored to offer, besides the scientific program of the conference, consisting in more than 65 high quality accepted papers, a rich and various series of social events, as usual for the spirit of this conference.

show abstract

Section: Numerical Experimentsmentioning

confidence: 99%

ECMS 2019 Proceedings edited by Mauro Iacono, Francesco Palmieri, Marco Gribaudo, Massimo Ficco

2019

View full text Add to dashboard Cite

show abstract

“…According to the JSQ policy the arriving job is routed to the server with minimum number of jobs; with the myopic policy the broker chooses such server, which minimizes the sojourn time for the arriving job. We used our own implementation of the JSQ and myopic policies, using the simulation framework Konovalov and Razumchik (2014); Konovalov (2014). The ties in the JSQ policy were broken according to the rule: if two or more servers have the same number of jobs (in the queue and in the server), choose the fastest server.…”

Section: Numerical Experimentsmentioning

confidence: 99%

Using Inter-Arrival Times For Scheduling In Non-Observable Queues

Konovalov¹,

Razumchik²

2017

ECMS 2017 Proceedings Edited by Zita Zoltay Paprika, Péter Horák, Kata Váradi, Péter Tamás Zwierczyk, Ágnes Vidovics-Dancs, Ján

Self Cite

View full text Add to dashboard Cite

In online dispatching systems, when there is no opportunity to observe the state of the systems' components, one may implement "blind" scheduling policies i.e. those which use incomplete/indirect observations of the system state or do not use any information at all. Here we deal with the well-known problem of scheduling in several non-observable parallel single server queues with single Poisson incoming flow, when the broker (scheduler) does not observe neither the current states of the queues and servers, nor the size of the incoming jobs. The only available information is the job size distribution, server's speeds and job's inter-arrival time distribution. For this problem setting it is known that if the scheduler can memorize the sequence of its previous decisions, then a deterministic policy is much better than the probabilistic policy (with respect to the job's mean waiting and mean sojourn time). But if the broker can memorize its decision, it is also very natural to assume that it can also memorize the time instants at which these decisions are made. In this paper we address the following question: can the deterministic policy be improved if the broker, in addition to decisions made, utilizes also the information about the lengths of time between the decisions? We give numerical evidence that it is indeed possible; the cases presented include three, five and nine parallel •|M|1 queues. We describe the simple new policy according to which, the broker's decisions are based on the estimates of the queue sizes. In most of the numerical experiments, the new policy outperformed the deterministic policy The relative gain may reach 10% in the case of the job's mean sojourn time and 50% in the case of the job's mean waiting time. 1 I.e. the broker does not have a queue to store the jobs. 2 As, for example, demonstrated in Anselmi and Gaujal (2010).

show abstract

“…Its formal description is based on the concept of communicating sequential processes introduced by C.A.R. Hoare (see Konovalov and Razumchik (2014); Konovalov (2014Konovalov ( , 2007).…”

Section: Obtaining the Best Possible Solutionmentioning

confidence: 99%

Simulation And Selection Of Efficient Decision Rules In Bank's Manual Underwriting Process

Konovalov¹,

Razumchik²

2016

ECMS 2016 Proceedings Edited by Thorsten Claus, Frank Herrmann, Michael Manitz, Oliver Rose

Self Cite

View full text Add to dashboard Cite

Bank's manual underwriting involves a group of underwriting inspectors, which perform a known set of procedures with the loan applications submitted by the borrowers, in order to determine the risk of providing a loan and eventually approve or disapprove it. Due to the fact that the evaluation process of applications must satisfy quality of service requirements usually set at legislator level and the bank resources are limited, one has to define such dispatching rules, that specify which application must be sent to which inspector and when in such a way that the requirements are met. This paper presents a case study of the application of "computer-aided scheduling" to the new problem of optimal management of applications, which is seen in the bank manual underwriting process. Here it is shown that the problem of optimal distribution of applications between the inspectors in the bank's manual underwriting can be represented as an optimal dispatching problem, commonly encountered in the distributed processing environment. We build the simulation model of the corresponding dispatching system and find best decision rule with the help of computational simulations. The realization of best decision-making is done by finding in a given set of dispatching rules the best one (either static or adaptive) for a given criterion and by estimating its parameters (if needed). By virtue of numerical examples it is shown how the quality of service requirements are met using different dispatching rules.

show abstract

Simulation of job allocation in distributed processing systems

Cited by 4 publications

References 3 publications

ECMS 2019 Proceedings edited by Mauro Iacono, Francesco Palmieri, Marco Gribaudo, Massimo Ficco

ECMS 2019 Proceedings edited by Mauro Iacono, Francesco Palmieri, Marco Gribaudo, Massimo Ficco

Using Inter-Arrival Times For Scheduling In Non-Observable Queues

Simulation And Selection Of Efficient Decision Rules In Bank's Manual Underwriting Process

Contact Info

Product

Resources

About