Describing and Forecasting Video Access Patterns
MetadataShow full item record
Citation (published version)Gursun, Gonca; Crovella, Mark; Matta, Ibrahim. "Describing and Forecasting Video Access Patterns", Technical Report BUCS-TR-2010-037, Computer Science Department, Boston University, November 10, 2010. [Available from: http://hdl.handle.net/2144/3811]
Computer systems are increasingly driven by workloads that reflect large-scale social behavior, such as rapid changes in the popularity of media items like videos. Capacity planners and system designers must plan for rapid, massive changes in workloads when such social behavior is a factor. In this paper we make two contributions intended to assist in the design and provisioning of such systems.We analyze an extensive dataset consisting of the daily access counts of hundreds of thousands of YouTube videos. In this dataset, we find that there are two types of videos: those that show rapid changes in popularity, and those that are consistently popular over long time periods. We call these two types rarely-accessed and frequently-accessed videos, respectively. We observe that most of the videos in our data set clearly fall in one of these two types. For each type of video we ask two questions: first, are there relatively simple models that can describe its daily access patterns? And second, can we use these simple models to predict the number of accesses that a video will have in the near future, as a tool for capacity planning? To answer these questions we develop two different frameworks for characterization and forecasting of access patterns. We show that for frequently-accessed videos, daily access patterns can be extracted via principal component analysis, and used efficiently for forecasting. For rarely-accessed videos, we demonstrate a clustering method that allows one to classify bursts of popularity and use those classifications for forecasting.