A relational approach to monitoring complex systems

Abstract

Monitoringis an essential part of many program development tools, and plays a central role in debugging, optimization, status reporting, and reconfiguration. Traditional monitoring techniques are inadequate when monitoring complex systems such as multiprocessors or distributed systems. A new approach is described in which a historical database forms the conceptual basis for the information processed by the monitor. This approach permits advances in specifying the low-level data collection, specifying the analysis of the collected data, performing the analysis, and displaying the results. Two prototype implementations demonstrate the feasibility of the approach.

Software approach to hazard detection using on-line analysis of safety constraints

“…Snodgrass has shown that a relational database query language can be used beneficially to specify queries that are evaluated against event streams such as are generated with on-line monitoring [23]. The significant difference between evaluating queries against a database and evaluating them against an event stream is that in the latter constraints must be evaluated against a conceptual database rather than an actual database.…”

“…It has been shown that the relational model is an appropriate formalism for the information processed by the monitor though earlier applications of this formalism to monitoring were primarily for performance evaluation [9,23]. In addition, prior approaches to monitoring were static, that is, they required that all constraints be known at compile time.…”

Run-time detection in parallel and distributed systems: application to safety-critical systems

“…Events may be extracted from files, received from sockets, or received as signals from an electro-mechanical sensor. When modeled in the context of our relational approach, an event is described as a database tuple [13]. All events of a given type are then members of the same relation, and relation membership is distinct (i.e., no tuple belongs to more than one relation.…”

“…Although similar to ours in its support of temporal and complex specifications, this work cannot be directly applied to run-time detection. Snodgrass [13] develops an information based approach to modeling program behavior that treats monitoring information (runtime data, states of processes, states of processors, messages, etc.) as relations.…”

Taking the step from meta-information to communication middleware in computational data streams

“…A false negative is a match that should have been made that is missed [9]. A reasonable window size depends strongly on the temporal ordering of the two input streams used in the join and on their relative synchrony [14]. For two event streams, each with a guaranteed partial order and containing an event for every logical timestep, minimal storage would be required.…”