LSP.9. Load Balancing in a Cloud

In a cloud infrastructure, typically data are sharded (or partitioned) and replicated for the purpose of increasing an overall cloud performance. A cloud additionally offers a dynamic scaling up by means of adding nodes and/or virtual machines and distributing a workload into multiple nodes. All of these techniques can be combined in order to balance a work done by every node and thus, possibly to increase an overall system's performance.

The aim of this topic is to: (1) analyze how the data partitioning and data allocation algorithms known from distributed file/operating systems can be adopted in the context of a cloud infrastructure, (2) develop data relocation and workload partitioning algorithms for balancing a workload, (3) extend the developed algorithms with scaling up policies (managing node additions/deletions), (4) experimentally evaluate the proposed algorithms w.r.t. balancing a workload and an overall system throughput.

Main Advisor at Poznan University of Technology (PUT)
Co-advisor at Technische Universit├Ąt Dresden (TUD)