I have a bachelor’s in electrical and computer engineering and a master’s in communications engineering. In between my bachelor’s and master’s, I worked at Nokia Siemens Networks for four years in various capacities. Currently, I am a PhD candidate at Aalto University in Finland. My research focus is on developing energy efficient multi-server queuing models for server farms while retaining performance SLAs.
Project: A Scheduling Problem in webMathematica
The scheduling policy currently used in the Wolfram|Alpha server cluster to route incoming requests is Round Robin. Each compute node in a cluster is running five instances of a webMathematica kernel. The Round Robin scheduler is supposed to evenly distribute the incoming requests across the compute nodes.
However, not all requests pass through the Round Robin scheduler. Often requests may have a recalculate phase, in which case, the subsequent requests should be sent to the compute node that served the original request, hence, bypassing the load balancer. Moreover, requests can be sent statically to a specific compute node.
With this setup, the Round Robin scheduler finds it hard to fulfill its intended task—load balancing. The aim of this project is to investigate this issue through simulation runs.