hide forever  |   hide once

Advanced Performance Model (APM)

Bandwidth Utilization Prediction Prototype

To improve the efficiency of resource utilization and scheduling of scientific data transfers on high-speed networks, we started a project on Advanced Performance Modeling (APM) that investigates and models a general-purpose network performance estimation framework. The performance prediction model and the framework will be helpful in optimizing the performance and utilization of fast networks as well as sharing resources with predictable performance for scientific collaborations, especially in data intensive applications. The prediction model estimates future network usage and the latency in using the network. Historical network performance information is used for throughput prediction without putting extra load on the resources by active measurement collection.
The details are describe in "Network Bandwidth Utilization Forecast Model on High Bandwidth Network", LBNL-6677E, 2014. (PDF)

Blue is the prediction of one day ahead of the current time. Red is the observed data of the same duration.	The prediction of one day ahead of the current time.

The bandwidth Utilization during 1 week before prediction

The bandwidth Utilization during 4 weeks before prediction

Related Documents

• Network Bandwidth Utilization Forecast Model on High Bandwidth Network, LBNL-6677E, 2014. (PDF)
• Relational Dynamic Bayesian Networks with Locally Exchangeable Measures, LBNL-6341E, 2013. (PDF)
• Estimating and Forecasting Network Traffic Performance based on Statistical Patterns Observed in SNMP data, MLDM2013. (PDF)

Disclaimers
© 2014 University of California
This web page is maintained by the APM project,
part of the Scientific Data Management Group
at the Lawrence Berkeley National Laboratory
Update: 12 Sep 2014