Corvil » White Papers
White Papers
- Nanosecond Latency Management
- Latency Management For Co-Location Trading
- Clock Synchronization
- Unified Latency Management
- Inter-Party Low Latency Management
- Managing Performance in Financial Trading Networks
- Measurement and Characterization of Latency in Trading Networks
- Analysis for Low Latency Trading Networks
- Traffic Conditions in Modern IP Networks
White Papers
Traffic Conditions in Modern IP Networks
It can be hard for a network manager to get a feeling for what the traffic on the network is really doing. After all, traffic is driven by machines that don’t operate on human timescales. Network congestion, when it occurs, can build up and disappear again in milliseconds. High-level monitoring tools alert us to a problem but leave us wondering: “What happened?” To understand the set of actions and reactions that drive performance, we resort to packet-by-packet analysis of massive packet captures. But in high-speed networks, this approach is not scalable: we find ourselves adrift in an ocean of data.
Here at Corvil we have worked with a number of customers over the past few years to understand the performance issues that they experience in the day-to-day life of their networks. In the process we have built up a large collection of traffic traces from operational networks. We use this data both to diagnose issues and to develop scalable analysis techniques that can be applied in the field.
The data we have gathered so far can be summarized in three words: diverse, disparate, and heterogeneous. Nevertheless, there are some common features that are surprising, and apparently ubiquitous. And we have developed some techniques that we feel are useful for analyzing them. This article summarizes what we have found so far.
