Summer 2011 ESCC/Internet2 Joint Techs

close
Use Internet2 SiteID

Already have an Internet2 SiteID?
Sign in here.

Internet2 SiteID

Scalable Network Measurement Analysis With Hadoop

Time 07/12/11 02:00PM-02:20PM

We do not host flash-based video files on our servers anymore. Please contact Web Support for further details about netcast videos.

Session Abstract

Available network performance data is accelerating due to the increased deployment of measurement infrastructure such as perfSONAR. Although there is a lot of research into offline analysis of this kind of data, most of these approaches do not address the systems and scalability issues. Applying industry best-practices, we use Hadoop to aggregate and filter the data to a set that fits into memory for further detailed analysis with R. This talk presents results from application of this method to a large set of perfSONAR measurements.

Speakers

Speaker Taghrid Samak Lawrence Berkeley National Laboratory

Presentation Media

media item thumbnail Scalable Network Measurement Analysis with Hadoop (pdf)

Speaker Taghrid Samak Lawrence Berkeley National Laboratory

Secondary tracks Research Partnership Performance / Measurement Advanced Network Services

Session Media

media item thumbnail Scalable Network Measurement Analysis With Hadoop Netcast Archive Available network performance data is accelerating due to the increased deployment of measurement infrastructure such as perfSONAR. Although there is a lot of research into offline analysis of this kind of data, most of these approaches do not address the systems and scalability issues. Applying industry best-practices, we use Hadoop to aggregate and filter the data to a set that fits into memory for further detailed analysis with R. This talk presents results from application of this method to a large set of perfSONAR measurements. media item thumbnail Scalable Network Measurement Analysis With Hadoop Netcast Archive Available network performance data is accelerating due to the increased deployment of measurement infrastructure such as perfSONAR. Although there is a lot of research into offline analysis of this kind of data, most of these approaches do not address the systems and scalability issues. Applying industry best-practices, we use Hadoop to aggregate and filter the data to a set that fits into memory for further detailed analysis with R. This talk presents results from application of this method to a large set of perfSONAR measurements.

gold Sponsors