2017 Technology Exchange

close
Use Internet2 SiteID

Already have an Internet2 SiteID?
Sign in here.

Internet2 SiteID

Your organization not listed? Create a local account to use Internet2 services.

Create SiteID

Wireless Performance Monitoring – From Infrastructure to Crowdsourcing

Time 10/16/17 02:30PM-03:00PM

Room Seacliff C

Session Abstract

Campus wireless networks have evolved into the primary network access for faculty, staff and students, eclipsing the campus wired network. Managing thousands of access points to serve a mobile device-centric audience is challenging now that wireless networks can no longer be “just good enough”. Instead, the community expects a highly available, ubiquitous (including residential areas) and secure network with seamless roaming. The twin pressures of 5 Ghz coverage and explosion of consumer and IoT wireless devices have led to a need to actively identify issues to improve the wireless experience for campus users.

This presentation will discuss the efforts we started in 2016 to meet the challenges that Duke has dealt with in expanding and securing wireless coverage to meet the increased demand from students, faculty and staff. In our experience, the wireless network is one of the most complex IT systems on campus, where issues could range from:

• Radio Frequency (RF) coverage and interference (also known as finding the appropriate balance between optimal performance, and user expectations)
• Increased complexity due to use of private IP addresses and port address translation for Internet access.
• Reliable and fast authentication mechanisms
• Ease of use
• Client-side issues and user reluctance to reporting issues
• Automated discovery of issues and faults in the wireless infrastructure

Rather than addressing the issues individually, Duke’s Office of Information Technology took a team approach with engineers for wireless, DevOps/automation, and log analysis working together to develop testing and statistical analysis of end user experience, testing methodologies for central IT (raspberry pi’s + trending), and Splunk dashboards to track issues reported by wireless LAN controllers and access points, and actively identify trouble spots independently of user complaints

This presentation will cover the development of the tools and techniques that we are using ranging from Boomerang javascript embedded in popular web pages to crowdsourcing network experience metrics, to placing raspberry pi’s as sensors in congested network segments, and employing third party tools to look at wireless traffic for issues. We will cover several before and after case studies that our approach helped us address. We will also discuss what led us to build our own monitoring framework, so that we could monitor as much of the end-to-end experience on wireless as possible.

Speakers

Speaker Richard Biever Duke University

Speaker Charles Francis Duke University

Speaker Stan Francis Duke University

Primary track Advanced Networking

Secondary tracks Information Security

gold Sponsors

bronze Sponsors