博士论文题目：Efficient and Low-Cost Data Management Using Multiple Public Clouds
导师: Franck MORVAN (email@example.com)
副导师: Shaoyi YIN (firstname.lastname@example.org)（信息学院校友）
Nowadays, huge amounts of data are produced by different sources (scientific observation, simulation, sensors, logs, social networks, finance). To manage these data, a cloud system approach allows reducing the cost in term of infrastructure either by purchasing a server comprised of low-cost commodity machines (Private Cloud) or by renting a service from a provider (Public Cloud) in pay-per-use. Public Clouds provide on demand resources and services. The tenants pay only the resources that they consume, while the provider has the opportunity to keep an economic gain due to resource consolidation.
For an application using Public Clouds, the overall objective is to minimize the bill, in condition that the query response time is acceptable. It is not easy to choose among multiple providers. First, different providers have different pricing policies, even though the bill always depends on consumed resources (i.e., CPU, RAM, disk and network). Second, the price of resources from a same provider may change over time depending on its commercial strategy. Finally, the resource requirements of different queries submitted by the same application may be various in terms of CPU, RAM, disk and network. For example, for a CPU-intensive query Q1, the provider P1 has the lowest cost; but for a network-intensive query Q2 from the same application, the provider P2 has the lowest cost. In this context, the aim of this thesis is to design and develop methods which allow (i) estimating the resource requirements of a query submitted by an application in function of pricing policies of cloud providers, (ii) selecting dynamically the cloud provider which is the least costly for executing the query.
图卢兹大学（Université de Toulouse）是法国著名综合性公立大学，建于1229年，是世界上最早的大学之一。其主要成员包括原图卢兹一大、图卢兹二大、图卢兹三大、法国国立应用科学院等17所高等院校，在多个领域拥有强大的师资力量。