Thursday, May 20, 2010

The changing traffic of R&E networks and cyber-infrastructure

[As I pointed out several months in my paper on the Future of R&E networks, traffic patterns are rapidly changing on all Internet networks including R&E networks. Increasingly traffic is becoming more local as these networks start to directly peer with large content, application and cloud providers at major IXs. A good example of this change in traffic pattern is the recent press release issued by the R&E network in the province of Ontario in Canada. They have seen a traffic volumes triple over the last 3 years. The overwhelming volume of the traffic is with over 40 content, cloud and application providers such as Google, Limelight etc. Only about 16% of the traffic goes to the national Internet backbone operated by CANARIE.

As cloud services for apps and research become more popular it is expected that this traffic volume will increase. For researchers clouds are still largely seen as a one to one replacement of physical clusters for virtual equivalents , but this is only the thin edge of the wedge for clouds. The real growth in clouds as a replacement for cyber-infrastructure will happen with next generation research applications that break down this limiting relationship between compute cycles and applications. This will be especially evident in the wireless sensor space and using mobile phone and pads as scientific instruments integrated with clouds. Please see my presentation on his subject at:

As Ed Lazowska points out in his one of blogs only a small percentage of cyber-infrastructure research applications need high performance computing facilities. Most of these applications can easily use clouds.

With the advent of the iphone and ipad a large percentage of traffic is now moving to wireless WiFi networks. Many public wifi networks now report that 50% of their traffic is from mobile data sources such as the Android, iphone and IPad. These volumes are expected to increase dramatically over the next few years. The proximity of content and application servers for wireless devices becomes more critical as data throughput can fall of f dramatically if there is any congestion on the wireless networks. Most likely a major future role for R&E networks will be to partner with various content and application providers to host their distributed infrastructure as close as possible to the end user. In addition to peering at small a number of IXs, hosting caching, content and cloud gateways/servers as near as possible to the institutions will significantly improve performance, particularly for wireless access. A major demand for lightpaths may not be to support research applications directly but to enable this widely distributed content, application and cloud infrastructure across a research networks service area. Access to this infrastructure will also enhance the performance and accessibility of education and research services offered by R&E institutions to users around the world.

ORON press release on tripling of traffic

Personal perspective on Future of R&E networks