A methodology for studying persistency aspects of internet flows

摘要: We focus in this paper on Internet flows, consider their contributions to the overall traffic per time unit or bin, and perform a multi-scale multi-protocol analysis explore persistency properties of those flows that contribute most (also known as "heavy hitters" "elephants"). Knowing features (or lack thereof) heavy hitters understanding underlying causes is crucial when developing engineering tools primarily optimizing system performance for elephant flows.The main difficulty studying available measurements are either too fine-grained large-scale studies (i.e., packet-level traces) coarse-grained extract detailed information necessary purpose at hand Netflow traces, SNMP). deal with problem by assuming have constant throughput through lifetime. then check validity assumption comparing our Netflow-derived findings against obtained from directly corresponding traces. By considering different aggregations (e.g., bin sizes between 1--10 minutes) flow abstractions raw IP pre-fix flows), varying definition what constitutes an "elephant", slicing protocols applications, we present methodology aspects exhibited flows. For example, find least once one unit) lifetimes tend show remarkable persistence be elephants much lifetimes, but certain aggregate exhibit more intricate properties.