Ana
Klimovic,
Electrical
Engineering
Department
Stanford
University
Data is exponentially being stored and processed in the cloud. Cloud computing promises high performance, cost-efficiency, and elasticity. However, to achieve these goals, cloud systems must provide each application with the right type and amount of storage and compute resources. Providing the right balance of resources is a major challenge due to the dynamic nature of cloud applications and the lack of flexibility in how today’s storage systems allocate resources. While cloud workload requirements vary widely over time and between applications, cloud servers are built with fixed ratios of storage and compute resources. This leads to over-allocation of resources, which increases cost.
In this talk, I will show how we can build cloud storage systems to achieve resource efficiency and high performance. The key insight is to decouple storage and compute resources, enabling any CPU core to access and share any storage device in a cloud facility that has available capacity and bandwidth. My work makes this approach practical in the context of modern cloud applications and modern storage hardware by addressing two fundamental requirements: 1) fast access to remote data and 2) intelligent, automatic control of storage resource allocation.
I will present ReFlex, a custom network-storage operating system that provides fast access to modern Flash storage over commodity cloud networks. ReFlex enables storage devices to be shared among multiple tenants with predictable performance. I will also present Pocket, a cloud storage service that builds on top of ReFlex. Pocket combines fast access to remote data with intelligent data placement and automatic resource allocation, enabling a new class of real-time analytics applications to run efficiently in the cloud.
Bio: Ana Klimovic is a Ph.D. candidate in the Electrical Engineering Department at Stanford University. Her research interests are in computer systems and computer architecture. Ana is particularly interested in designing and implementing high performance, resource-efficient computer systems for cloud computing.
As part of her research, she has collaborated with companies such as Facebook, Microsoft, and IBM. Before coming to Stanford, Ana earned her Bachelor’s degree in Engineering Science at the University of Toronto. She is a Microsoft Research Ph.D. Fellow, Stanford Graduate Fellow, and Accel Innovation Scholar.