About SAND

The SAND project coordinates network monitoring and diagnostics from over a hundred application-level measurement points and aggregates them into a measurement archive. This curated measurement archive provides access to historical and quasi-real-time network performance data, allowing for higher-level diagnastics and analyses.

SAND supports and collaborates with long-running activities such as the OSG-Networking area and the WLCG Throughput working group. The data collected by each of these projects goes into the SAND Network Monitoring Archive (SAND-NMA).

SAND maintains a map of CC* sites running perfSONAR; to allow your site to show up, follow these directions.

IRIS-HEP Summer 2020 Fellowship Presentations

Sample packet loss graph

Packet Loss Tracking

Sample packet loss graph from the SAND archive

View Dashboard

Sample TCP throughput graph

Throughput Tracking

SAND’s archives record historical single-stream TCP across many links

View Dashboard

Traceroute Dashboard

Traceroute Tracking

SAND’s archives record historical traceroutes throughout many links

View Dashboard

MEPHi TRACER (traceroute explorer)

MEPHi TRACER

Interactive Traceroute explorer

Open App

PS-Dash App

PS-Dash App

OSG-LHC/WLCG network metrics summary, site rankings, and highlighted problem pairs

Open App

SAND Network Monitoring Archive

The SAND-NMA aims to aggregate as much network monitoring tests results as feasible, providing a knowledgebase for debugging networks and performing data analytics. This archive consists of a data collector (polling measurement endpoints), a message bus (which endpoints can also push data into directly), an archive, and a visualization platform, as pictured here:

SAND-NMA Architecture

SAND-NMA Architecture

Currently, we are recording network measurements from perfSONAR, HTCondor file transfer, and and XRootD TCP transfers.

A few metrics about our service:

  • SAND users: Engaged BNL (ATLAS), AGLT2 (ATLAS), MWT2 (ATLAS), UCSD (CMS), Nebraska (CMS), Wisconsin (CMS), MIT (CMS), and Purdue (CMS) to identify and resolve network issues
  • Number of data sources: 323 distinct perfSONAR endpoints
  • Number of HTCondor submit hosts reporting network measurements: 16
  • Number of distinct measurements in the ElasticSearch database: over 16 billion, from February 28, 2018 to present

Contact us

We can be reached via our public Google Group.