You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 11 Next »

Description

perfSONAR is an open-source, modular and flexible network monitoring architecture. Key features are:

- Multi-domain: perfSONAR provides a complete picture of network performance across multiple domains allowing NOC and PERT engineers to seamlessly view, analyse and diagnose network behaviour across the entire network path.

 - Scalable: The service is collaboratively delivered by GÉANT and participating NRENs, allowing those NRENs to install, deploy and manage their own monitoring installation.

perfSONAR (Performance Service-Oriented Network monitoring ARchitecture ) is a protocol, architecture and set of tools which have been developed specifically to address the challenge of monitoring network performance in a multi-domain environment and is the only architecture  capable of performing this. perfSONAR is produced by an international collaboration between organisations in Europe, North America and Brazil with the collaborating organisations using the perfSONAR concept to develop their own service to their own requirements.
 

PerfSONAR deployments based on the perfSONAR software product operate by installing measurement points (MPs) across the networks. These MPs run the PerfSONAR client software and allow users to perform a range of measurements across the network.

The MP is a small server (or virtual server) which is connected directly to a suitable network interface in the Point of Presence.  Each MP can run software to measure the following metrics to aid the analysis and correction of issues;

- Bandwidth
- Delay
- Jitter 
- Route Tracing

Once installed, the MP is registered with the GEANT PerfSONAR team and it is included within the network of measurement points available to PerfSONAR members.
 
At any time, any approved PerfSONAR engineer can access the GEANT PerfSONAR infrastructure and request a variety of ad hoc measurements between any two MPs within the environment.  Additionally, it is possible to schedule regular test measurements between MPs to enable performance monitoring and trend analysis.  Requests for regular measurements are made via the GEANT PerfSONAR team and the results of these tests are stored centrally.
 

All testing and analysis functions are undertaken through a straightforward web based User Interface with no client software requirements. This allows the use of mobile devices to access real-time performance measurements.

By using dedicated Measurement Point equipment PerfSONAR is able to provide consistent measurement across multiple domains which may be using different technology platforms and networking technologies.  This enables PerfSONAR to be hardware and technology independent. This platform independence also allows for PerfSONAR to be used during and after network migrations to assist in the deployment of new technologies within the NREN. 

The network information is extracted and collated from each MP and then presented in a standardised, universally-usable format through the perfSONAR user interface. Network engineers can access network metrics for their own network or for any other perfSONAR MP across the GÉANT Service Area, providing the complete picture of the network’s status across different domains.

The new converged perfSONAR aims at providing a performance monitoring and diagnostics infrastructure for NOC and PERT engineers in the GEANT service area, helping them to identify network performance issues, leverage the people and tools necessary to improve their end-to-end experience and verify SLA performance parameters to ensure the performance levels desire by the users of the NRENs are met.

NRENs already perform performance measurements within their domains and collect various data. This data will be made available when deploying the perfSONAR software in the perfSONAR service. In this way engineers will be able to use and combine the readily available performance information to provide a consistent end-to-end performance picture, delivering functionality that is useful to existing NREN workflows. 

The priorities will be to deliver a lightweight and stable set of tools for domains to easily and effectively manage their deployment, so that the overall quality of the multi-domain service provided is met.

Target users of a monitoring service based on perfSONAR are primarily members of PERTs and NREN NOCs who support end users of IP connectivity services provided over the GÉANT Service Area and who need to view performance data or perform tests (e.g. for event planning, research or operational needs). However, the ultimate beneficiary of the service is the end user. By providing the multi-domain monitoring and diagnostics infrastructure as a service to network performance teams and NOCs, performance issues for end users can collaboratively be identified, prevented and solved.

Initial perfSONAR  service use cases is multi-Domain network performance troubleshooting. In the next future two additional use cases will be inlcuded:

- Circuit monitoring (dynamically and statically provisioned).
- Multi-Domain performance service quality verification.
In this way, participating NREN NOC and PERT engineers, using perfSONAR MDM will be able to identify network problems, measure network metrics and monitor circuits across multiple domains.
Benefits for the NOC and PERT engineers will be:
- Accurate, quick and efficient detection of bottlenecks, outages and performance issues where problems span multiple domains.
- Assistance in the proactive identification and prevention of problems before service disruption occurs, helping to provide fast, reliable and uninterrupted network communication.
This will affect the end users (i.e. researchers, scientists, students) by improving their overall perception of network performance and of the connectivity services, providing, in particular:
- Quicker solutions to network outages
- Fewer disruptions
- Fewer performance issues

KPIs

Kpi nameRAGKPI RAG
FoobarGreenGREEN
DingbatGreenGREEN
Broken KPIRedRED
Another onepurpleYELLOW
  • No labels