"Telco-grade reliability" is a term typically associated with expensive, internally redundant hardware components. Conventional wisdom in the telecommunications industry is that the failure rate of commodity x86-based hardware is too high to support 99.99% or better availability. Due to a number of high-profile scalability and fault-tolerance failures of web systems, the prevailing view is that web-style- architectures are fundamentally unscalable and failures are to-be-expected.
We have tackled the problem of delivering highly-scalable, robust "telco-grade reliability" using commodity x86-based hardware. Tellme answers over 2 million calls per day using this system. In the process, we've redefined the traditional web architecture and have arrived at a generalizable approach to large-scale commodity systems.
In this talk, we will describe the architecture we've come to term "the appliance model". We will discuss its implications for scalability, fault-tolerance, recovery, persistent-data management, software design, deployment and configuration management, and testing.
About the speaker:
Brad joined Tellme as the company's fifth employee. He developed Tellme's original VoiceXML interpreter and served as Platform Architect responsible for overseeing our system scalability and VoiceXML interpreter development initiatives. He serves on the W3C standards committee as an editor of VoiceXML 2.0 and a lead architect for their Multimodal initiatives. As Tellme's Director of Engineering for Service Production, Brad leads the architecture, development, and quality assurance of all client solutions for Tellme.
Before joining Tellme, Brad helped architect Netscape Netcenter's personalization and cobranding infrastructure. Brad is graduate of the Massachusetts Institute of Technology where he earned a Master's of Engineering and a B.S. in Computer Science.
Email: bwporter at-sign yahoo.com