Abstract: A hierarchical registration architecture manages and organizes the collection of performance information, such as statistics and tracing, for an extensible operating system of a computer. The registration architecture, or registry, comprises a multi-linked tree data structure within a main memory for accessing the performance information. Writer entities register their intent to collect and store performance information in the registry by creating objects, via novel API calls, as nodes organized within the tree structure. Each object node of the registry is named according to a convention that identifies the type of performance data collected by that node. Each object node further represents a single data item having a single data type for collecting the performance and a reference to the actual storage location of its collected performance information. Information storage preferably occurs in globally-shared memory so that any software entity can access the information.

Abstract: A dynamic adaptive sampling process for providing performance analysis on a computer system; the sampling process comprising the steps of: generating a profile tree comprising a hierarchy of one or more nodes, each node in the hierarchy comprises an array of counters, with each counter mapped to an associated range of memory locations in the computer memory. The process then receives a program counter value and searches the profile tree according to the hierarchy of nodes to locate an associated counter. The associated counter maps to a corresponding memory range of the finest partition which includes the memory location identified by the program counter value. The sampling process then detects whether the located counter exceeds a threshold value, wherein if the located counter exceeds the threshold value, then the sampling process further comprises modifying the profile tree. Wherein if the located counter does not exceed the threshold value, then incrementing that located counter.