Markov Decision Processes: Discrete Stochastic Dynamic Programming
Markov Decision Processes: Discrete Stochastic Dynamic Programming
Neuro-Dynamic Programming
Technical Update: Least-Squares Temporal Difference Learning
Machine Learning
Resource overbooking and application profiling in shared hosting platforms
ACM SIGOPS Operating Systems Review - OSDI '02: Proceedings of the 5th symposium on Operating systems design and implementation
Least-squares policy iteration
The Journal of Machine Learning Research
Classifier-Based Policy Representation
ICMLA '08 Proceedings of the 2008 Seventh International Conference on Machine Learning and Applications
Approximate dynamic programming: lessons from the field
Proceedings of the 40th Conference on Winter Simulation
Natural actor-critic algorithms
Automatica (Journal of IFAC)
| Hi-index | 0.00 |
In this work, we optimize the admission policy of application deployment requests submitted to data centers. Data centers are typically comprised of many physical servers. However, their resources are limited, and occasionally demand can be higher than what the system can handle, resulting with lost opportunities. Since different requests typically have different revenue margins and resource requirements, the decision whether to admit a deployment, made on time of submission, is not trivial. We use the Markov Decision Process (MDP) framework to model this problem, and draw upon the Approximate Dynamic Programming (ADP) paradigm to devise optimized admission policies. We resort to approximate methods because typical data centers are too large to solve by standard methods. We show that our algorithms achieve substantial revenue improvements, and they are scalable to large centers.