PROJECT TITLE :

Every Timestamp Counts: Accurate Tracking of Network Latencies Using Reconcilable Difference Aggregator - 2018

ABSTRACT:

User-facing services deployed in data centers must respond quickly to user actions. The measurement of network latencies is of paramount importance. Recently, a brand new family of compact data structures has been proposed to estimate one-means latencies. In order to attain scalability, these new methods depend upon timestamp aggregation. Unfortunately, this approach suffers from serious accuracy problems within the presence of packet loss and reordering, given that a single lost or out-of-order packet might invalidate a large number of aggregated samples. During this Project, we have a tendency to unify the matter to detect lost and reordered packets within the set reconciliation framework. Although the set reconciliation approach and the info structures for aggregating packet timestamps are previously known, the mix of these 2 principles is novel. We have a tendency to gift a area-efficient synopsis referred to as reconcilable difference aggregator (RDA). RDA maximizes the proportion of useful packets for latency measurement by mapping packets to multiple banks and repairing aggregated samples that are broken by lost and reordered packets. RDA simultaneously obtains the typical and the standard deviation of the latency. We have a tendency to provide a formal guarantee of the performance and derive optimized parameters. We have a tendency to further style and implement a user-space passive latency measurement system that addresses sensible issues of integrating RDA into the network stack. Our in depth analysis shows that compared with existing methods, our approach improves the relative error of the typical latency estimation in 10-fifteen orders of magnitude, and also the relative error of the quality deviation in 0.5-6 orders of magnitude.