Balakrishna J. Prabhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Balakrishna J. Prabhu has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
- Olivier Brun
LAAS-CNRS, University of Toulouse, CNRS, Toulouse, France
9 shared publications - Jacques Resing
Department of Mathematics and Computer Science, Eindhoven University of Technology, Eindhoven, The Netherlands
3 shared publications - Rudesindo Núñez-Queija
Univ Amsterdam, Amsterdam, Netherlands
3 shared publications - Dieter Fiems
Department of Telecommunications and Information Processing, Ghent University, Gent, Belgium
2 shared publications - Matthieu Jonckheere
Universidad de Buenos Aires and CONICET
2 shared publications - Akshay Mete
Texas A&M University System, College Station, United States
1 shared publication
Latest Publications
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 37 (7) , 1647-1656
Performance Bounds for Priority-Based Stochastic Coflow Scheduling
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 37 (7) , 1647-1656
Performance Bounds for Priority-Based Stochastic Coflow Scheduling
- Open Access
Robust Online Coflow Scheduling from Predictions
In this paper, we introduce an online scheduling framework for minimizing the total weighted average completion time of coflows, in scenarios where both flow sizes and release times are unknown. Instead of relying on exact flow sizes, the framework leverages potentially inaccurate predictions. At each decision point, it selects a subset of coflows by solving a Minimum Unscheduled Weight Problem (MUWP), and then applies a variant of the Sincronia algorithm-adapted to operate directly on the predicted sizes. We prove that the proposed approach achieves provably strong competitive guarantees, and extensive simulations demonstrate that it outperforms baseline methods both on average and in worst-case instances.
Open AccessRobust Online Coflow Scheduling from Predictions
In this paper, we introduce an online scheduling framework for minimizing the total weighted average completion time of coflows, in scenarios where both flow sizes and release times are unknown. Instead of relying on exact flow sizes, the framework leverages potentially inaccurate predictions. At each decision point, it selects a subset of coflows by solving a Minimum Unscheduled Weight Problem (MUWP), and then applies a variant of the Sincronia algorithm-adapted to operate directly on the predicted sizes. We prove that the proposed approach achieves provably strong competitive guarantees, and extensive simulations demonstrate that it outperforms baseline methods both on average and in worst-case instances.
- 21 January 2025
- Published by Springer Nature
- p. 14-23
Strategic Routing in Heterogeneous Discriminatory Processor Sharing Queues
- 21 January 2025
- Published by Springer Nature
- p. 14-23
Strategic Routing in Heterogeneous Discriminatory Processor Sharing Queues
- 25 May 2024
- Published by Springer Nature
- p. 118-137
Data Summarization for Federated Learning
- Open Access
- 21 May 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Modeling and Performance Evaluation of Computing Systems
- Vol. 9 (3) , 1-32
Partial Server Pooling in Delay Systems
We consider a strategic resource pooling interaction between two service providers, each of which is modelled as a multi-server queue. The traditional full resource pooling mechanism, while improving the overall system performance measure, may not always benefit the providers individually. For rational service providers, there will therefore be no incentive to form a stable coalition. We address this issue by proposing three partial sharing mechanisms: two of these, Partial Pooling with Dedicated Servers and Partial Pooling with Repacking, allow each provider to designate a certain number of servers to potentially serve jobs of the other provider, while the third mechanism, Differential Queue Length, maintains an upper bound on the difference between occupancies of the two queues. We determine the probability of waiting for each of the three mechanisms and show through several numerical examples that their Pareto frontiers are non-empty. In particular, this shows that the proposed partial pooling mechanisms are individually beneficial to both providers and incentivize coalition formation in all situations. Finally, we apply bargaining theory to determine the operating point on the Pareto frontier.
- 21 May 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Modeling and Performance Evaluation of Computing Systems
- Vol. 9 (3) , 1-32
Open AccessPartial Server Pooling in Delay Systems
We consider a strategic resource pooling interaction between two service providers, each of which is modelled as a multi-server queue. The traditional full resource pooling mechanism, while improving the overall system performance measure, may not always benefit the providers individually. For rational service providers, there will therefore be no incentive to form a stable coalition. We address this issue by proposing three partial sharing mechanisms: two of these, Partial Pooling with Dedicated Servers and Partial Pooling with Repacking, allow each provider to designate a certain number of servers to potentially serve jobs of the other provider, while the third mechanism, Differential Queue Length, maintains an upper bound on the difference between occupancies of the two queues. We determine the probability of waiting for each of the three mechanisms and show through several numerical examples that their Pareto frontiers are non-empty. In particular, this shows that the proposed partial pooling mechanisms are individually beneficial to both providers and incentivize coalition formation in all situations. Finally, we apply bargaining theory to determine the operating point on the Pareto frontier.
- 3 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cloud Computing
- Vol. 12 (2) , 644-658
Weighted Scheduling of Time-Sensitive Coflows
Datacenter networks commonly facilitate the transmission of data in distributed computing frameworks through coflows, which are collections of parallel flows associated with a common task. Most of the existing research has concentrated on scheduling coflows to minimize the time required for their completion, i.e., to optimize the average dispatch rate of coflows in the network fabric. Nevertheless, modern applications often produce coflows that are specifically intended for online services and mission-crucial computational tasks, necessitating adherence to specific deadlines for their completion. In this paper, we introduce $\mathtt {WDCoflow}$ , a new algorithm to maximize the weighted number of coflows that complete before their deadline. By combining a dynamic programming algorithm along with parallel inequalities, our heuristic solution performs at once coflow admission control and coflow prioritization, imposing a $\sigma$ -order on the set of coflows. With extensive simulation, we demonstrate the effectiveness of our algorithm in improving up to $3\times$ more coflows that meet their deadline in comparison the best SoA solution, namely $\mathtt {CS\rm{-}MHA}$ . Furthermore, when weights are used to differentiate coflow classes, $\mathtt {WDCoflow}$ is able to improve the admission per class up to $4\times$ , while increasing the average weighted coflow admission rate.
- 3 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cloud Computing
- Vol. 12 (2) , 644-658
Weighted Scheduling of Time-Sensitive Coflows
Datacenter networks commonly facilitate the transmission of data in distributed computing frameworks through coflows, which are collections of parallel flows associated with a common task. Most of the existing research has concentrated on scheduling coflows to minimize the time required for their completion, i.e., to optimize the average dispatch rate of coflows in the network fabric. Nevertheless, modern applications often produce coflows that are specifically intended for online services and mission-crucial computational tasks, necessitating adherence to specific deadlines for their completion. In this paper, we introduce $\mathtt {WDCoflow}$ , a new algorithm to maximize the weighted number of coflows that complete before their deadline. By combining a dynamic programming algorithm along with parallel inequalities, our heuristic solution performs at once coflow admission control and coflow prioritization, imposing a $\sigma$ -order on the set of coflows. With extensive simulation, we demonstrate the effectiveness of our algorithm in improving up to $3\times$ more coflows that meet their deadline in comparison the best SoA solution, namely $\mathtt {CS\rm{-}MHA}$ . Furthermore, when weights are used to differentiate coflow classes, $\mathtt {WDCoflow}$ is able to improve the admission per class up to $4\times$ , while increasing the average weighted coflow admission rate.
- Open Access
- 5 April 2023
- Published by Elsevier in Indagationes Mathematicae
- Vol. 34 (5) , 990-1013
Markovian queues with Poisson control
- 5 April 2023
- Published by Elsevier in Indagationes Mathematicae
- Vol. 34 (5) , 990-1013
Open AccessMarkovian queues with Poisson control
- Open Access
Markovian queues with Poisson control
We investigate Markovian queues that are examined by a controller at random times determined by a Poisson process. Upon examination, the controller sets the service speed to be equal to the minimum of the current number of customers in the queue and a certain maximum service speed; this service speed prevails until the next examination time. We study the resulting two-dimensional Markov process of queue length and server speed, in particular two regimes with time scale separation, specifically for infinitely frequent and infinitely long examination times. In the intermediate regime the analysis proves to be extremely challenging. To gain further insight into the model dynamics we then analyse two variants of the model in which the controller is just an observer and does not change the speed of the server.
Open AccessMarkovian queues with Poisson control
We investigate Markovian queues that are examined by a controller at random times determined by a Poisson process. Upon examination, the controller sets the service speed to be equal to the minimum of the current number of customers in the queue and a certain maximum service speed; this service speed prevails until the next examination time. We study the resulting two-dimensional Markov process of queue length and server speed, in particular two regimes with time scale separation, specifically for infinitely frequent and infinitely long examination times. In the intermediate regime the analysis proves to be extremely challenging. To gain further insight into the model dynamics we then analyse two variants of the model in which the controller is just an observer and does not change the speed of the server.
- 1 April 2022
- Published by Springer Nature in Queueing Systems
- Vol. 100 (3-4) , 233-235
Queueing models with service speed adaptations at arrival instants of an external observer
- 1 April 2022
- Published by Springer Nature in Queueing Systems
- Vol. 100 (3-4) , 233-235
Queueing models with service speed adaptations at arrival instants of an external observer
- 4 May 2021
- Published by Elsevier in Performance Evaluation
- Vol. 149-150, 102216
Editorial: Valuetools 2019
- 4 May 2021
- Published by Elsevier in Performance Evaluation
- Vol. 149-150, 102216
Editorial: Valuetools 2019