*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
TITLE: No Need for Guesswork: Traffic-oblivious Resource Management Policies for Multi-server systems
SPEAKER: Varun Gupta, Faculty Candidate
ABSTRACT:
Will it snow next week? Who will the next Super Bowl? Will my call center receive more than ten thousand calls tomorrow?
Uncertainty in traffic demand significantly impacts the operations decisions of multi-server systems (e.g., call centers, data centers, cloud platforms). For example, 1. Load Balancing: How should the load be distributed among a collection of heterogeneous servers? Should a request be dispatched to an idle slow server, or should it wait for a busy fast server?
2. Capacity Provisioning: How many servers should be commissioned to meet the demand? When should more servers be commissioned? When should existing servers be decommissioned to avoid waste?
While demand forecasting is the de facto approach to overcome uncertainty, we ask: Are there simple load balancing and capacity provisioning policies that work well without predicting the traffic demand? By obviating the need for traffic analysis, not only are such policies simpler to deploy, they are also robust to forecasting errors. In the talk, I will present traffic-oblivious load balancing and capacity provisioning policies that answer the above questions affirmatively.
