Software-defined latency monitoringin data center networks

Abstract

Data center network operators have to continually monitorpath latency to quickly detect and re-route traffic away from high-delaypath segments. Existing latency monitoring techniques in data centersrely on either (1) actively sending probes from end-hosts, which isrestricted in some cases and can only measure end-to-end latencies, or(2) passively capturing and aggregating traffic on network devices, whichrequires hardware modifications.In this work, we explore another opportunity for network path latencymonitoring, enabled by software-defined networking. We propose SLAM,a latency monitoring framework that dynamically sends specific probepackets to trigger control messages from the first and last switches of apath to a centralized controller. SLAM then estimates the latency distributionalong a path based on the arrival timestamps of the controlmessages at the controller. Our experiments show that the latency distributionsestimated by SLAM are sufficiently accurate to enable thedetection of latency spikes and the selection of low-latency paths in adata center