Software Defined Measurement for Data Centers

Traffic measurement plays an important role in data centers. Many data center management tasks ranging from performance diagnosis, anomaly detection to traffic accounting and engineering all rely on accurate and on-time measurement of the traffic from different tenants and applications. For example, to perform traffic engineering, we need to correctly detect large flow aggregates and pick better routes for these flows. To reduce the latency of partition/aggregate workloads, we need to quickly identify short traffic bursts (e.g., Incast). These measurement tasks vary in three aspects: which traffic to measure, where to measure, and when to measure.