SCS Undergraduate Thesis Topics
|Jiaqi Tan||Priya Narasimhan||BlackSheep: Inferring White-Box Application Behavior Using Black-Box Techniques|
We describe and evaluate a new technique for diagnosing performance problems in distributed systems in a scalable manner by exploiting and analyzing only local (i.e. intra-node) black-box system metrics, and inferring white-box application behavior. We study the novel method of correlating white-box application event logs with black-box system metrics to gain insight into the behavior of a distributed system, and validate our approach through experiments on the Hadoop open-source implementation of Google's Map/Reduce distributed programming model. We inject failures and real performance problems gathered from failure data recorded in Hadoop's bug database.