HP Labs Technical Reports

The Value of a Systematic Approach to Measurement and Analysis: An ISP Case Study

Ramanathan, Srinivas; Perry, Ed

HPL-98-171

Keyword(s): internet performance; service management

Abstract: Internet Service Providers (ISPs) are looking to differentiate their services (e.g., Web, Email, News) by offering quality of service guarantees to their subscribers. However, the absence of effective proactive monitoring and the inability to diagnose problems in a timely manner have meant that ISPs are unable to offer guarantees for the availability and performance of their services. Instead, ISPs mainly offer network-level quality guarantees for availability and delay, that too restricted to a single administrative domain. In this paper, we describe a system for monitoring and diagnosis of ISP services that can be used by ISPs to offer service- level guarantees. Service availability and performance measures included in this system document conformance to service-level guarantees. Coupled with these high- level health reports are measures of infrastructure services such as the Domain Name Service (DNS) and Network File Service (NFS) that can be used for problem diagnosis. To further assist with diagnosis and to provide information for capacity planning, measures of server and network utilization are included in the monitoring system. Measures of the availability and usage of an ISP ga ss Points of Presence (POPs) complete a comprehensive monitoring suite for ISPs. The integration of measurements into a service-oriented monitoring system, as described herein represents a significant advance in the state- of-the-practice for ISPs. We present experimental results obtained from a real-world ISP system that demonstrate the practical utility and effectiveness of the monitoring system in detecting and diagnosing problems in ISP environments.

22 Pages

Back to Index