TimeMap - tool for latency and jitter monitoring

Gravatar for Gloria Vuagnin
TrackPlenary
DescriptionMonitoring a network without knowing how the data flows is like reporting traffic conditions an an highway ignoring transit times and stops-and-go situations: just of little use if you want to predict how your journey will be. Transit time and stop-and-go translate into data packets latency and delivery time jitters on network and Timemap is the tool we developed to monitor them inside the GEANT backbone, a tool which can be / hopefully will be implemented also inside other national and international backbones. Services and Applications in strict need to know latency and jitter conditions to work correctly are out there and used daily, and some of them (like the very low latency high quality videoconferencing services in use in Performing Arts education and real time interactivity) need this monitoring to ensure a quick and reliable debugging when problems happen, too. We will show how we implemented the monitoring tool, directly talking to backbone routers to get the data, storing them and displaying and analyzing them; all using standard and open source well established components, with nearly zero "new developed software" to ensure an easy sustainability of the tool. Furthermore, since we started to collect historic data, we discovered interesting events which happen on the backbone, which were quite difficult to detect before, expanding our knowledge of the network itself. TimeMap is already taking data from the whole GEANT backbone since quite some time now, and as such also Anomaly Detection using machine learning techniques is now also working quite well. Besides the GEANT backbone, TimeMap can be used inside any other network backbone, national, regional, local. Adopting it will create a complet end to end monitoring system which can truly enable safe and reliable operations for any service which requires real time data transfer in a timely way.

Presentation documents

All talks