Eric Bauer

End-to-end availability considerations for services over IMS

Lucent IP Multimedia Subsystem (IMS) architecture provides an open framework for multimedia applications that support blended text, voice, and video services. Providing objectives for service downtime and failure rates, which reflect the end customer perspective, can be used to drive corrective and preventive action that better satisfy end customer expectations. However, the frameworks and techniques to do this on an end-to-end basis for network solutions of the complexity and size of IMS are not well established. In this paper we address the questions of specifying, estimating, and verifying the end-to-end availability for services over IMS. The purpose of this paper is to present a framework for establishing availability requirements and service failure rate metrics, and for performing “end-to-end” service downtime analysis. The framework can be used to guide network design and evaluate end-to-end performance in the field.

Modeling planned downtime

In general, service providers have only placed system availability requirements on equipment providers for unplanned system downtime. However, reliability expectations of service providers are shifting in a manner that places requirements on equipment providers to manage all system downtime (i.e., planned system downtime as well as unplanned system downtime) rather than focusing solely on guaranteeing unplanned system downtime. Furthermore, service providers continuously request improvements in total system availability (e.g., requiring 99.999% system availability, which translates to 5.25 minutes of total system downtime per year). While numerous models exist for modeling unplanned system downtime, there are no equivalent models for planned downtime. This paper presents a planned downtime event taxonomy, characterizes the planned downtime for each category in the taxonomy, and incorporates the results into a model for predicting the total planned downtime associated with a system. It then uses the model on an example system, demonstrating the capability of the model in assessing various system design alternatives.

Framework for availability characterization by analyzing outage durations

Telecommunications service providers that maintain their own networks tend to be very good at diagnosing and repairing equipment failures. Careful analysis of service providers own trouble ticket data can reveal actual failure rates, failure modes, failure footprints, and outage duration information. As trouble tickets inherently include both small capacity-loss events and fast switchover — which may ordinarily be excluded from TL9000 calculations — as well as planned/scheduled downtime, this offers a holistic view of product availability. This paper presents a taxonomy for analyzing these downtime events and guidance on both extracting salient product attributes from the data and on using the data to calibrate both unplanned and planned downtime models.

Appendix C: Software Reliability Growth Models

This appendix contains sections titled: Software Characteristic Models Nonhomogeneous Poisson Process Models

Appendix A: System Reliability Report Outline

This appendix contains sections titled: Executive Summary Reliability Requirements Unplanned Downtime Model and Results Annex A??????-??????Reliability Definitions Annex B??????-??????References Annex C??????-??????Markov Model State-Transition Diagrams ]]>