Buffer-Based High-Coverage and Low-Overhead Request Event Monitoring in the Cloud

Abstract

Request latency directly affects the performance of modern cloud applications. Due to various causes in hosts and networks, can suffer from latency anomalies (), which may violate the Service-Level Agreement. However, existing performance monitoring tools have incomplete coverage and inconsistent semantics for monitoring and cannot accurately diagnose . This paper presents, a high-coverage and low-overhead event monitoring system, which monitors <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">buffers</i> to capture most -related abnormal events with consistent -level semantics in the end-to-end datapath of . First, models the datapath of as a buffer chain and defines events based on three properties of buffers, so as to <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">end-to-end monitor</i> the root causes of . Then, to achieve <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">consistent semantics</i> for captured events, designs a -level semantics injection mechanism to make events captured in networks have the victim requests’ ID. Finally, offloads the semantics operations and event collection in software to SmartNICs for <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">low CPU overhead</i> . We have implemented on commodity SmartNICs and programmable switches. Evaluation results show that can diagnose 98% with <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$&lt;$</tex-math> </inline-formula> 0.08% network bandwidth overhead and 0.6% application throughput decline.