Jonatha Anselmi
Research scientist
Hi, I'm a tenured researcher (chargé de recherche ‐ CRCN) at the National Institute for Research in Digital Science and Technology (Inria)
having the luck of working in the POLARIS team.
I got my PhD degree in computer engineering from
Politecnico di Milano (Italy) in 2009,
where I also received a B.Sc. (2003) and a M.Sc. (2005) cum laudem degree.
After my PhD experience, I've joined Inria in Grenoble as a postdoc (20092010) and then the Basque Center for Applied Mathematics (BCAM) in Bilbao (20102013). In 2014, I've joined Inria as a tenured researcher working with the CQFD team in Bordeaux until 2018.
Among others,
I've held visiting positions at
IBM T.J. Watson
and Caltech.
jonatha.anselmi ∈ inria.fr
+33 4 57 42 16 15
LIG  Bâtiment IMAG (Bureau 445)
700 avenue Centrale, 38400 St Martin d'Hères, France
Research
I'm interested in the broad field of decision making under uncertainty, at the intersection of applied mathematics, computer science and engineering. Drawing mostly on methods from applied probability and operations research, I'm looking for highlyscalable algorithms that minimize congestion and operational costs of largescale distributed systems. I like to use Markov processes to model the dynamics of cloud networks, and then study delays and power consumption costs in limiting regimes of practical interest.
Keywords: queueing theory, reinforcement learning, Markov decision processes, load balancing, autoscaling, power consumption minimization, cloud systems.
Selected recent publications
 J. Anselmi
Asynchronous Load Balancing and Autoscaling: Meanfield Limit and Optimal Design,
IEEE/ACM Transactions on Networking [10.1109/TNET.2024.3368130]
 J. Anselmi, B. Gaujal, L.S. Rebuffi
Reinforcement Learning in a Birth and Death Process: Breaking the Dependence on the State Space
, NeurIPS '22: Proceedings of the 36th International Conference on Neural Information Processing Systems, 2022
 J. Anselmi, N. Walton
Stability and Optimization of Speculative Queueing Networks,
IEEE/ACM Transactions on Networking, vol. 30, no. 2, pp. 911922, April 2022
 J. Anselmi, F. Dufour
PowerofdChoices with Memory: Fluid Limit and Optimality,
Mathematics of Operations Research, 45, 3, 862888, 2020
 J. Anselmi Combining SizeBased Load Balancing with RoundRobin for Scalable Low Latency, IEEE Transactions on Parallel and Distributed Systems, vol. 31, no. 4, pp. 886896, 2020
Students
A PhD position is available on the themes of
i) reinforcement learning in structured Markov decision processes
and
ii) load balancing and autoscaling.
If you are a student interested in the broad field of decision making under uncertainty,
with a solid background in computer science or applied math, please contact me.
PhD students:
 LouisSebastien Rebuffi (with B. Gaujal). Subject: "Reinforcement Learning Algorithms for Controlled Queueing Systems" [thesis defended on December 2023].
Interns:
[Reinforcement learning] Bel HouariDurand Sehane [0208/2024]
[Autoscaling and load balancing] Jules BURGAT [0208/2023], Mingming DAI [0208/2023], Dorian BUFFIÈRE [0309/2022]
Publications
My publications on [Google scholar] [hal.science] [dblp.org]
Refereed journal papers
 J. Anselmi, B. Gaujal, L.S. Rebuffi
Learning Optimal Admission Control in Partially Observable Queueing Networks, Queueing Systems (to appear)
 J. Anselmi
Asynchronous Load Balancing and Autoscaling: Meanfield Limit and Optimal Design,
IEEE/ACM Transactions on Networking [10.1109/TNET.2024.3368130]
 J. Anselmi, J. Doncel
Load Balancing with JobSize Testing: Performance Improvement or Degradation?,
ACM Transactions on Modeling and Performance Evaluation of Computing Systems, 9, 2, Article 8, 2024, 27 pages.
 F. Filippini, J. Anselmi, D. Ardagna, B Gaujal A Stochastic Approach for Scheduling AI Training Jobs in GPUbased Systems , IEEE Transactions on Cloud Computing, vol. 12, no. 01, pp. 5369, 2024.
 J. Anselmi, N. Walton Stability and Optimization of Speculative Queueing Networks, IEEE/ACM Transactions on Networking, vol. 30, no. 2, pp. 911922, April 2022.
 J. Anselmi Replication vs speculation for load balancing, Queueing Systems, 100, 389–391 (2022).
 J. Anselmi, B. Gaujal, L.S. Rebuffi Optimal Speed Profile of a DVFS Processor under Soft Deadlines, Performance Evaluation, Vol. 152, 2021, 102245 (also accepted at IFIP Performance '21) [video]
 J. Anselmi, F. Dufour PowerofdChoices with Memory: Fluid Limit and Optimality, Mathematics of Operations Research, 45, 3, 862888, 2020.
 J. Anselmi Combining SizeBased Load Balancing with RoundRobin for Scalable Low Latency, IEEE Transactions on Parallel and Distributed Systems, vol. 31, no. 4, pp. 886896, 2020
 J. Anselmi, J. Doncel Asymptotically Optimal SizeInterval Task Assignments, IEEE Transactions on Parallel and Distributed Systems, vol. 30, no. 11, pp. 24222433, 2019
 J. Anselmi, F. Dufour, T. PrietoRumeau Computable approximations for average Markov decision processes in continuoustime, Journal of Applied Probability, 55(2), 571592, 2018
 J. Anselmi Asymptotically optimal openloop load balancing, Queueing Systems, Vol. 87, 34, pp. 245267, 2017
 J. Anselmi, D. Ardagna, J. C.S. Lui, A. Wierman, Y. Xu, Z. Yang The economics of the cloud: price competition and congestion, ACM Transactions on Modeling and Performance Evaluation of Computing Systems, Vol. 2, 4 (18), 2017
 J. Anselmi, F. Dufour, T. PrietoRumeau Computable approximations for continuoustime Markov decision processes on Borel spaces based on empirical measures, Journal of Mathematical Analysis and Applications, 443 (2), 1323–1361, 2016
 J. Anselmi, N.S. Walton Decentralized Proportional Load Balancing, SIAM Journal on Applied Mathematics, 76(1), 391410, 2016
 J. Anselmi, B. Gaujal, T. Nesti Control of parallel nonobservable queues: asymptotic equivalence and optimality of periodic policies, Stochastic Systems, Vol. 5 No. 1, (2015), 120145
 M. Passacantando, J. Anselmi, D. Ardagna Generalized Nash Equilibria for PlatformasaService Clouds, European Journal of Operational Research, 236(1):326339, 2014
 J. Anselmi, B. Gaujal Efficiency of simulation in monotone hyperstable queueing networks, Queueing Systems, 76(1):5172, 2014
 J. Anselmi, B. D'Auria, N. Walton Closed queueing networks under congestion: nonbottleneck independence and bottleneck convergence, Mathematics of Operations Research, 38:3, 469491, 2013
 J. Anselmi, G. Casale HeavyTraffic Revenue Maximization in Parallel Multiclass Queues, Perform. Eval., 70(10): 806821 (2013)
 T. Radivojevic, J. Anselmi, E. Scalas Ergodic transition in a simple model of the continuous double auction Plos one 9(2): e88095.
 J. Anselmi, B. Gaujal The 'Price of Forgetting' in parallel and nonobservable queues, Perform. Eval., 68(12): 12911311 (2011)
 J. Anselmi, U. Ayesta, A. Wierman Competition yields efficiency in load balancing games, Perform. Eval., 68(11): 9861001 (2011)
 J. Anselmi, I.M. Verloop Energyaware capacity scaling in virtualized environments with performance guarantees, Perform. Eval., 68(11): 12071221 (2011)
 J. Anselmi, P. Cremonesi A unified framework for the bottleneck analysis of multiclass queueing networks, Perform. Eval., 67(4): 218234 (2010)
 J. Anselmi, Y. Lu, M. Sharma, M. S. Squillante Improved approximations for the Erlang loss model, Queueing Systems, 63(14): 217239 (2009)
Refereed conference papers
 J. Anselmi, B. Gaujal, L.S. Rebuffi Reinforcement Learning in a Birth and Death Process: Breaking the Dependence on the State Space , NeurIPS '22: Proceedings of the 36th International Conference on Neural Information Processing Systems, 2022.

J. Anselmi, B. Gaujal
Energy Optimal Activation of Processors for the Execution of a Single Task with Unknown Size, IEEE MASCOTS 2022, 30th International Symposium on the Modeling, Analysis, and Simulation of Computer and Telecommunication Systems, October 1820, 2022 Nice, France
Best Paper Award  H. Zhang, F. Dufour, J. Anselmi, D. Laneuville, A. Nègre Piecewise Optimal Trajectories of Observer for BearingsOnly Tracking of Maneuvering Target, 2018 IEEE Aerospace Conference, 2018, pp. 17
 H. Zhang, F. Dufour, J. Anselmi, D. Laneuville, A. Nègre Piecewise Optimal Trajectories of Observer for BearingsOnly Tracking by Quantization, 20th International Conference on Information Fusion (Fusion), 2017, pp. 17
 J. Anselmi, B. D'Auria, N. Walton A ProcessorSharing Heuristic for Multipath Congestion Control, Proc. of the 51st Annual Allerton Conference on Communication, Control, and Computing, 2013, pp. 19
 J. Anselmi, B. Gaujal Optimal Routing in Parallel, nonObservable Queues and the Price of Anarchy Revisited, 22nd International Teletraffic Congress, ITC22, IEEE, 2010
 J. Anselmi, B. Gaujal Performance Evaluation of a Work Stealing Algorithm for Streaming Applications, 13th Int. Conf. On Principle Of DIstributed Systems (OPODIS), 2009, Springer LNCS, Vol. 5923/2009, pp. 1832.
 J. Anselmi, E. Amaldi, P. Cremonesi On the Consolidation of Datacenters with Performance Constraints, 5th Int. Conf. on the Quality of Software Architectures (QoSA), 2009, Springer LNCS, Vol. 5581/2009, pp. 163176.

J. Anselmi A New Framework Supporting the Bottleneck Analysis of Multiclass Queueing Networks, ACM VALUETOOLS 08: Third Int. Conf. on Performance Evaluation Methodologies and Tools, October 2024, 2008, Athens, Greece.
 J. Anselmi, P. Cremonesi Bounding the Performance of BCMP Networks with LoadDependent Stations, IEEE MASCOTS 2008, 16th Annual Meeting of the IEEE Int. Symp. on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems, Baltimore, Maryland, USA, 810 September 2008.
 J. Anselmi, P. Cremonesi Exact Asymptotic Analysis of Closed BCMP Networks with a Common Bottleneck, ASMTA'08: The 15th Int. Conf. on Analytical and Stochastic Modelling Techniques and Applications, Springer LNCS, Nicosia, Cyprus, 46 June 2008
 J. Anselmi, E. Amaldi, P. Cremonesi Service Consolidation with EndtoEnd Response Time Constraints, SEAA'08: Euromicro Conference on Service Engineering and Advanced Applications, Italy, IEEE, Parma, 35 September 2008
 J. Anselmi, G. Casale, P. Cremonesi Approximate Solution of Multiclass Queueing Networks with Region Constraints, IEEE MASCOTS 2007, 15th IEEE Int. Symp. on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems, Istanbul, Turkey, October 2007.
Workshop papers and technical reports
 J. Anselmi, D. Ardagna, J. C.S. Lui, A. Wierman, Y. Xu, Z. Yang The economics of the cloud: price competition and congestion, SIGecom Exchanges 13(1): 5863 (2014). Also appeared in: SIGMETRICS Performance Evaluation Review 41(4): 4749 (2014)
 M. Larrañaga, J. Anselmi, U. Ayesta, P. Jacko, A. Romo Optimization techniques applied to railway systems, HAL technical report, 2013
 J. Anselmi, B. Gaujal On the efficiency of perfect simulation in monotone queueing networks, Perform. Eval. Rev., Vol. 39, No. 2, pp. 5658, 2011 [extended version]
 J. Anselmi, B. Gaujal The Price of Anarchy in Parallel Queues Revisited, Proc. of ACM SIGMETRICS 2010, pp. 353354. Poster presentation.
 J. Anselmi, Y. Lu, M. Sharma, M. S. Squillante Improved approximations for stochastic loss networks, Perform. Eval. Rev., Vol. 37, No. 2, pp. 4547, 2009.
 J. Anselmi, P. Cremonesi Bounding the Partition Function of BCMP Multiclass Queueing Networks BWWQT'09: Belarusian Winter Workshop on Queueing Theory, Minsk, Belarus, January 2009 [slides]
 J. Anselmi, G. Casale, P. Cremonesi A PopulationMix Driven Approximation for Queueing Networks with Finite Capacity Regions, BWWQT'07: Belarusian Winter Workshop on Queueing Theory, Grodno, Belarus, January 2007 [extended version]
 J. Anselmi, D. Ardagna, P. Cremonesi A QoSBased Selection Approach of Autonomic Grid Services, HPDC/SOCP '07: ACM Proc. of the 2007 workshop on Serviceoriented computing performance: aspects, issues, and approaches, Monterey, CA, USA, 2007.
Teaching
 Probabilités et simulation ‐ PolyTech Grenoble, INFO4
 Évaluation de performances ‐ PolyTech Grenoble, INFO4
If you're a student, you may find the R code snippets written in class or for homework useful.