Big data analytic behind simple TCP traffic curves?



        The three curves are too simple, aren't they? Actually it is very difficult to compute. Network traffic visualization is to reveal network protocol properties for the underlying packets. For the TCP protocol, the most important ones are:
        1)the amount of traffic, measuring the volume of communication
        2)end-to-end latency, measuring time to send a packet
        3)retransmission rate, measuring the effectiveness to send a packet.
        Believe or not it is very hard to make them right. Netflow/sFlow can only measure the TCP traffic at about 70% accuracy since it samples the switch traffic, let alone calculating the TCP latency and the TCP retransmission rate.
        Even harder, the TCP traffic in an enterprise environment has two directions LAN (west-to-east), which flows internally, and WAN (north-to-south), that accesses the Internet. The traffic behavior of LAN and WAN is different and it requires more advanced analysis to distinguish them. That is the reason that most curves you see are per HOST based.
        After working out this problem for years and relying on big-data analytic power recently, we finally put the TCP traffic visualization pieces together. You can see we could not only accurately measure the traffic properties but also effectively distinguishing the two's. What a marvelous achievement in TCP traffic visualization!
© 2013-2019 IDO-NET All rights reserved.