Real-time monitoring and control in a multi-nodal, multi-tiered distributed system

摘要

A method involves starting multiple nodes; executing production workloads, made up of transactions; running a Monitoring and Control Program (MCP) on each node, wherein at least one MCP acts as a controlling MCP; monitoring in real-time individual transaction execution, the monitoring involving collecting parametric information regarding execution of the transactions on a node and transaction basis, the parametric information including at least an identification of each specific transaction, each node involved in processing those specific transactions, and at least one timing value related to the processing of each specific transaction; generating a transaction table including the collected parametric information; identifying whether there is degraded performance within the computer system, based upon throughput per unit time for all individual transactions; and, when degraded performance is identified, running lower level analysis only for particular nodes and particular transactions associated with the degraded performance.