A scalable SNMP-based distibuted monitoring system for heterogeneous network computing
Proceedings of the 2000 ACM/IEEE conference on Supercomputing
Low-overhead message tracking for distributed messaging
Proceedings of the ACM/IFIP/USENIX 2006 International Conference on Middleware
Hi-index | 0.00 |
Steady increases in the adoption of web services for automating critical business processes or activities demand that certain Quality of Service (QoS) metrics are met. Common approaches to QoS monitoring are primarily based on a centralized model that we found to be inadequate in large-scale and high-volume computing environments, such as Intel's Assembly and Test manufacturing facilities. The main reasons are: scalability, issues related to practical deployment, and reconfigurability. In this paper, we will present a peer-to-peer approach that works by transparently turning a web service into a potential QoS monitor for all activities that flow through it. A lightweight middleware component, residing within the web service implementation, provides this transparency. The middleware component collects QoS metrics from a web service during activity processing and embeds the metrics within the business data payload. This allows the peers of that service to receive QoS metrics, apply specific monitoring rules, or take specific actions as the activity flows from one web service to the next. This decentralized pattern allows activities of different complexity and length to be monitored seamlessly. We will discuss our experience in implementing and using this technique at Intel's high-volume Assembly and Test manufacturing environment that span across multiple continents and geographic regions.