9:15-10:45 Session 1: Markov Chains, Markov Decision Processes
| Time |
Details |
| 9:15 - 9:25 |
Opening remarks |
| 9:25 - 9:45 |
Markov chain aggregation with error bounds on transient distributions Fabian Michel and Markus Siegle |
| 9:45 - 10:05 |
A lazy abstraction algorithm for Markov decision processes: theory and initial evaluation Dániel Szekeres, Kristóf Marussy and Istvan Majzik |
| 10:05 - 10:25 |
Deep reinforcement learning for weakly coupled MDP’s with continuous actions Francisco Robledo, Urtzi Ayesta and Konstantin Avrachenkov |
| 10:25 - 10:45 |
Strong aggregation in the stochastic matching model with random discipline Jean-Michel Fourneau and Moyi Yang |
10:45-11:10 Break
11:10-13:15 Session 2: Queues
| Time |
Details |
| 11:10 - 11:30 |
Queueing analysis of an ensemble machine learning system Keishin Tsutsumi, Tuan Phung-Duc and Hong-Linh Truong |
| 11:30 - 11:50 |
Revenue management for parallel services with fully observable queues Caitlin Vanden Bussche, Arnaud Devos, Sabine Wittevrongel and Dieter Fiems |
| 11:50 - 12:10 |
Optimizing load balancing for heterogeneous M/M/c/K server clusters in the stationary mean-field regime Illes Horvath and Marton Meszaros |
| 12:10 - 12:30 |
An algebraic proof of the relation of Markov fluid queue and QBD processes Peter Buchholz, Andras Meszaros and Miklos Telek |
| 12:30 - 12:50 |
Stability of the multiserver job queuing model with infinite resources Adityo Anggraito, Diletta Olliaro, Andrea Marin and Marco Ajmone Marsan |
| 12:50 - 13:10 |
Optimal allocation of tasks to networked computing facilities Vincenzo Mancuso, Paolo Castagno, Leonardo Badia, Matteo Sereno and Marco Ajmone Marsan |
13:15-14:45 Lunch
14:45-16:15 Joint Keynote Talk ASMTA and EPEW
Abstract:
Through an array of service models such as SaaS, PaaS, and IaaS, cloud computing has become an indispensable part of modern business operations, offering a wide range of benefits driving its rapid adoption, including versatility, scalability, and security. However, cost optimization still remains a difficult challenge specifically for right-sizing, i.e., choosing the instance configurations that best suit the workload. In this talk, I will overview past and current efforts to address this issue using software performance modeling and optimization, including recent initiatives at applying research results into industry.