2015/04/02 by Jonathan Curtis Beard, Beard, Jonathan C., Roger D. Chamberlain +1
Computer Science · #Advanced Database Systems and Queries #Cloud Computing and Resource Management #Distributed and Parallel Computing Systems #FOS: Computer and information sciences #Parallel Computing and Optimization Techniques #Performance (cs.PF)
paper · pdf · doi:10.48550/arxiv.1504.00591
openalex publication_date 2015/04/02 · openalex created_date 2022/10/02 · openalex updated_date 2026/07/28
Stream processing is a compute paradigm that promises safe and efficient\nparallelism. Modern big-data problems are often well suited for stream\nprocessing's throughput-oriented nature. Realization of efficient stream\nprocessing requires monitoring and optimization of multiple communications\nlinks. Most techniques to optimize these links use queueing network models or\nnetwork flow models, which require some idea of the actual execution rate of\neach independent compute kernel within the system. What we want to know is how\nfast can each kernel process data independent of other communicating kernels.\nThis is known as the "service rate" of the kernel within the queueing\nliterature. Current approaches to divining service rates are static. Modern\nworkloads, however, are often dynamic. Shared cloud systems also present\napplications with highly dynamic execution environments (multiple users,\nhardware migration, etc.). It is therefore desirable to continuously re-tune an\napplication during run time (online) in response to changing conditions. Our\napproach enables online service rate monitoring under most conditions,\nobviating the need for reliance on steady state predictions for what are\nprobably non-steady state phenomena. First, some of the difficulties associated\nwith online service rate determination are examined. Second, the algorithm to\napproximate the online non-blocking service rate is described. Lastly, the\nalgorithm is implemented within the open source RaftLib framework for\nvalidation using a simple microbenchmark as well as two full streaming\napplications.\n