TICP: Transport Information collection Protocol

Overview

Project details

Papers

Talks

Implementation

People involved

Project details (from the beginning to more recent)

TICP (version 0)	A collector machine sends probes to a set of information sources, which answer by sending back their report packets containing their information. The collect session is a succession of rounds. In the first round, the collector probes all sources. In other rounds, it asks sources whose reports where lost in previous rounds to retransmit. This behavior in rounds is meant to wait for transitory network congestions to disappear from one round to another and to absorb the excessive delay that some reports may experience. To control the rate of sending probes, TICP uses a window-based report-clocked congestion control similar to the TCP one. New probes are sent only if the number of expected reports is less than the congestion window size. TICP adapts the congestion window to the observed loss rate. It uses for this two algorithms: Slow start and Congestion Avoidance. For more details, please read this paper.
TICP with clustering of information sources	We improve TICP by adding a new mechanism that clusters information sources. This mechanism aims at smoothing the variation of the congestion control parameters and at probing together sources behind the same bottleneck. Our clustering mechanism is based upon the Global Network Positioning (GNP) Internet coordinate system. The Internet is modeled as a 2-dimensional Euclidean space. Each information source has a pair of coordinates in this space. The Euclidean distance between any two network nodes gives an estimate of the Round-Trip Time. We divide the plane into disjoint square areas of equal sizes and we define a cluster as being a set of information sources whose representing points are located in the same square area. The central square is the one whose centre is the point representing the collector. TICP probes clusters from the nearest to the most distant starting from the central cluster and following a spiral trajectory. Inside a cluster, sources are probed in a random order. Probing clusters We refer to this paper for more details and for a performance evaluation of the clustering solution.
TICP for large size information collection	In this version, TICP supports collecting large quantities of data from each information source. The collector is now able to gather unlimited number of packets from each data source. We have compared this version to a collection application using parallel TCP connections to retrieve packets from sources. We proved that TICP has shorter collect session duration because it avoids transitory phase in each individual TCP connection. More explications on this version will be given in a future publication.
TICP with collection delegation	With a well-chosen number of proxy collectors, TICP collect session duration can decrease dramatically. How to choose proxy collectors? How many proxy collectors are optimal? What is the behavior of a proxy collector? We are currently answering these questions and results will be provided in a future publication.