Discovering likely invariants of distributed transaction systems for autonomic system management

Abstract

Large amount of monitoring data can be collected from distributed systems as the observables to analyze system behaviors. However, without reasonable models to characterize systems, we can hardly interpret such monitoring data effectively for system management. In this paper, a new concept named flow intensity is introduced to measure the intensity with which internal monitoring data reacts to the volume of user requests in distributed transaction systems. We propose a novel approach to automatically model and search relationships between the flow intensities measured at various points across the system. If the modeled relationships hold all the time, they are regarded as invariants of the underlying system. Experimental results from a real system demonstrate that such invariants widely exist in distributed transaction systems. Further we discuss how such invariants can be used to characterize complex systems and support autonomic system management.

Search-based inference of polynomial metamorphic relations

“…For example, Ernst et al [21] developed a tool named Daikon to discover program invariants for supporting program evolution. Jiang et al [32] proposed a novel technique to automatically model and search relationships between the flow intensities that can be regarded as invariants. Csallner et al [16] proposed to infer invariants using dynamic symbolic execution.…”

A query language and runtime tool for evaluating behavior of multi-tier servers

“…Related work in the area of fault diagnosis has focused on four approaches: (i) using statistical correlations [1,2,4,9,24], (ii) using models [18,23], (iii) using specialized languages [11,12,16], and (iv) using fine-grained profiling [5,7].…”

“…Pinpoint [2] and Magpie [1] are statistical tools for fault detection in component-based Internet service. Another approach is to use invariants -those metric correlations that hold in a variety of conditions as the correctness measure [9]. Cohen et al [4] correlate system metrics to high-level states to find the root cause of faults.…”

Semantic-Driven Model Composition for Accurate Anomaly Diagnosis

“…Gaussian Mixture Models: Gaussian Mixture Models (GMM), a clustering algorithm, is used to capture the statistical correlation between pairs of metrics [3,4].…”

“…In contrast, recent research [3] [4] [5] [6] [7] [8] has investigated developing statistical models for automatic fault detection. These techniques derive probabilistic relationships, called functional invariants or correlations, between metrics captured at different points across the system.…”