You have downloaded this file 0 times in the last 24 hours, limit is 50.
Your file downloads total 0 in the last 24 hours, limit is 100.

Creating a Scalable Architecture for Internet Measurement Download


Internet congestion has become a serious problem for the U.S. Research and Education community, with many users experiencing difficulty obtaining adequate performance from the Internet. One key factor contributing to this situation is the lack of any mechanism to diagnose end-to-end performance problems. Further, there is little market incentive for Internet providers to address performance issues.

We propose to build an automated measurement infrastructure on top of the existing Internet. The architecture for this infrastructure would be similar to that of other large scale Internet services, such as the Domain Name System, the Network Time Protocol, and the USENET news system. Each of these applications succeeds in scaling to the full Internet because they are structured in a hierarchical, distributed fashion, allowing their infrastructure to scalably grow without significant central administration. Similarly, we envision a distributed National Internet Measurement Infrastructure (NIMI), consisting of probe machines situated throughout the Internet. These probes can be used to monitor end-to-end (probe-to-probe) performance of the Internet to assess how well it continues to meet the needs of groups such as the U.S. Research and Education community. The information provided by the probes can also be used to localize faults to particular pieces of the national Internet infrastructure.

Although we cannot require sites to conduct a uniform set of measurements and to publish the results in a uniform way, we envision providing market-based incentives for doing so. For example, sites purchasing high-speed network service could include, as part of their contracts, requirements concerning the presence and use of NIMI platforms. This would allow them to assure that they are indeed receiving the performance for which they contracted. A large federal agency such as the NSF could further stipulate as part of its service contracts that the service provider conduct and publish NIMI measurements to contribute to the national-level view of the Internet as a whole. A second type of incentive comes from the diagnostic capability of the infrastructure as a whole: the ability to detect both current and long-term performance degradation of portions of the Internet. This should motivate providers to cooperate with each other in setting up a global mesh of NIMI probes. Further, since a NIMI measurement requires cooperation at both ends (the two measurement platforms conducting the end-to-end measurement), sites that publish their results could refuse to cooperate in conducting measurements with sites that do not publish results. This would give non-publishing providers a strong incentive to publish, if they wish to continue to reap the diagnostic benefits of NIMI. Such publicity will provide a strong incentive for providers to raise the performance and quality of their services.

The goal of our proposal is to pilot a technology demonstrating the necessary scaling properties. The initial deployment will serve as a template for others to replicate the technology throughout the Internet. Once the seeds of the NIMI technology are planted, the infrastructure should prove sufficiently valuable to both Internet providers and end-users that it becomes in everyone's interest to support its further growth, without requiring continued support from any single funding source. Thus, the long-term deployment of the NIMI technology will largely be funded by the Internet community itself.

The probe architecture will be patterned after Vern Paxson's {\em Network Probe Daemon} (NPD), which was recently used to perform the first large-scale measurements of end-to-end Internet behavior (routing and connection dynamics). The NPD experiment included 37 participating sites and measured more than 1,000 distinct Internet paths. Its success provides us with both the confidence that the measurement approach is feasible, and the experience of how to overcome the difficulties associated with such a broad scale task. We give an overview of NPD in the next section.

One important component of the NIMI is the capability of monitoring the Internet's large-scale performance. To this end, we propose as part of our efforts to develop a framework for producing a {\em national Internet profile}---a collection of measurements from which we can assess the health of the national Internet as a whole, over time. This profile will be constructed from measurements produced by NIMI probes and archived at the PSC. One type of "state of the Internet" bulletin will be daily summaries of performance metrics among the mesh of probes (Internet "traffic reports" a la "Eye in the Sky" updates for automotive traffic). This reduced data will be made publicly available via a WWW interface integrated with NLANR.

In addition to "state of the Internet" bulletins, the archives will serve as an invaluable resource for Internet research. By maintaining a continuous baseline of traffic data, researchers can study how the Internet evolves and infer cause-and-effect relationships in a manner that is difficult to undertake today.

Submitted On:
29 Aug 2012
File Size:
40.94 Kb
File Author:
A. Adams
Submitted On:
29 Aug 2012
Total Votes:0
Be the first to comment! Please sign in or register.