2nd Performance Management Workshop

Over the last three months, we have been running network performance tests involving perfSONAR hosts in Europe, Brazil and USA. We have
also compared results with tests run on a test bed with four hosts connected to a switch.

One special case of our tests involved testing the new Bela link that connects Europe (through Portugal) to Brazil at 100G. We were aiming to identify routing problems, and to obtain maximal throughput with a minimum of retransmits. We tested the effect of different congestion of algorithms, socket length and MTU.

We report sustained throughput from Netherlands to Fortaleza above 96Gbps. We present cases of limited performance between 100G hosts and try to identify the reasons for low performance.

The talk will present the tests and measurements made on 100 Gigabit network infrastructure using TCP/IP. It will cover three domains: between end sites in the UK and Italy, between Latin America and Europe using the EllaLink Submarine Cable, and transfers between Australia to Europe to support SKA using both a path via the US and a path on the CAE1 link.

Performance Measurement Platform explores performance of the GÉANT backbone while experiencing perfSONAR on small nodes. In our work we explore the possibilities of using machine learning algorithms on perfSONAR collected data in order to detect network anomalies.

Timemap is a real-time monitoring tool to collect, visualize and analyze latency and jitter. Timemap is collecting data from the Géant backbone devices since April 2021. Recent Timemap updates introduce support for streaming ML anomaly detection. In this talk we will introduce the algorithms adopted in Timemap, the usability improvement enabled by the new feature and we will discuss some detection performance aspects.

NetSage is an open privacy-aware network measurement, analysis, and visualization service designed to address the needs of today's international networks. Modern science is increasingly data-driven and collaborative in nature, producing petabytes of data that can be shared by tens to thousands of scientists all over the world. NetSage services provide an unprecedented combination of passive and active measurements, including SNMP data, flow data, and tstat-based traffic analysis, perfSONAR, and longitudinal network performance data visualization. Several new applications of the NetSage framework are being explored for campus use, upcoming HL-LHC data challenges, and as the basis for other monitoring platforms such as ESnet's Stardust. This talk will explain how NetSage operates, its current applications, and plans for future development.