Automatic Failure Diagnosis Support in Distributed Large-Scale Software Systems Based on Timing Behavior Anomaly Correlation

Marwede, Nina; Röhr, Matthias; Hoorn, André van; Hasselbring, Wilhelm

doi:10.1109/csmr.2009.15

Cited by 37 publications

(15 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MonitorRank consistently outperformed the basic heuristics that were employed by LinkedIn's monitoring team and current state-of-the-art anomaly correlation algorithms [23]. In terms of mean average precision, which quantifies the goodness of root cause ranking, MonitorRank yields 26% to 51% more predictive power than any other technique we tried.…”

Section: Introductionmentioning

confidence: 84%

Root cause detection in a service-oriented architecture

Kim

Sumbaly

Shah

2013

Proceedings of the ACM SIGMETRICS/international Conference on Measurement and Modeling of Computer Systems

View full text Add to dashboard Cite

Large-scale websites are predominantly built as a service-oriented architecture. Here, services are specialized for a certain task, run on multiple machines, and communicate with each other to serve a user's request. An anomalous change in a metric of one service can propagate to other services during this communication, resulting in overall degradation of the request. As any such degradation is revenue impacting, maintaining correct functionality is of paramount concern: it is important to find the root cause of any anomaly as quickly as possible. This is challenging because there are numerous metrics or sensors for a given service, and a modern website is usually composed of hundreds of services running on thousands of machines in multiple data centers. This paper introduces MonitorRank, an algorithm that can reduce the time, domain knowledge, and human effort required to find the root causes of anomalies in such service-oriented architectures. In the event of an anomaly, MonitorRank provides a ranked order list of possible root causes for monitoring teams to investigate. MonitorRank uses the historical and current time-series metrics of each sensor as its input, along with the call graph generated between sensors to build an unsupervised model for ranking. Experiments on real production outage data from LinkedIn, one of the largest online social networks, shows a 26% to 51% improvement in mean average precision in finding root causes compared to baseline and current state-of-the-art methods.

show abstract

Section: Introductionmentioning

confidence: 84%

Root cause detection in a service-oriented architecture

Kim

Sumbaly

Shah

2013

Proceedings of the ACM SIGMETRICS/international Conference on Measurement and Modeling of Computer Systems

View full text Add to dashboard Cite

show abstract

“…RanCorr will contribute to the diagnosis perspective by providing information on problem root causes. The static visualization of RanCorr results described in [3] will then be reimplemented based on ExplorViz's interactive visualization capabilities.…”

Section: ) θPadmentioning

confidence: 99%

“…2) RanCorr: Based on component anomaly scores -e.g., provided by ΘPAD -RanCorr [3] localizes the root cause of problems by correlating the anomaly scores with architectural information in form of calling dependencies. RanCorr will contribute to the diagnosis perspective by providing information on problem root causes.…”

Section: ) θPadmentioning

confidence: 99%

“…The control center is designed for extensibility by providing a plug-in architecture enabling the integration into visualization perspectives. With respect to visualization of software runtime behavior, our previous work has focused on providing different statically generated levels of visualization ( [3], [4]). This can become tedious, if the software landscape provides thousands of those levels.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Towards a Dependability Control Center for Large Software Landscapes (Short Paper)

Fittkau

Hoorn²,

Hasselbring

2014

2014 Tenth European Dependable Computing Conference

Self Cite

View full text Add to dashboard Cite

Manual management of dependability while operating large software systems -including failure detection, diagnosis, repair, and prevention activities -is time-consuming and error-prone. Various automatic approaches supporting these activities have been proposed, e.g., to detect and diagnose performance degradation effects caused by software aging and to execute reactive or proactive rejuvenation actions. However, users often mistrust fully-automatic dependability management approaches due to a lack of control over the change actions conducted to the business-critical software landscape. Building trust for automatic systems is challenging.In this paper, we present our envisioned control center for a semi-automatic management of large software landscapes, featured by a graphical user interface including interactive system visualizations. The control center will provide a reusable platform for integrating techniques for dependability management, including monitoring, and analyzing a system's dependability during production as well as for planning and executing reactive or proactive change actions to the software landscape.

show abstract

“…Live visualization with ExplorViz is scalable [6] and elastic in cloud environments [28]. Monitoring may provide runtime models [23] for system comprehension [9], trace visualization [4], architecture conformance checks [11], and a landscape control center [12] with performance anomaly detection [3,24]. New perspectives on employing virtual reality [8] and physical models [7] are further explored.…”

Section: Icpementioning

confidence: 99%

Microservices for Scalability

Hasselbring

2016

Proceedings of the 7th ACM/SPEC on International Conference on Performance Engineering

Self Cite

View full text Add to dashboard Cite

Microservice architectures provide small services that may be deployed and scaled independently of each other, and may employ different middleware stacks for their implementation. Microservice architectures emphasize transaction-less coordination between services, with explicit acceptance of eventual consistency. Polyglott persistence in this context means that the individual microservices may employ multiple data storage technologies. Microservice architectures are "cloud native" allowing for automated and rapid elasticity. Fault-tolerance mechanisms achieve that failures of individual mircroservices do not affect other services thanks to container isolation. Since services can fail at any time, it is important to be able to detect the failures quickly and, if possible, automatically restore services. Essential for success in such a setting is advanced monitoring.In this keynote, I discuss how mircoservices support scalability for both, runtime performance and development performance, via polyglott persistence, eventual consistency, loose coupling, open source frameworks, and continuous monitoring for elastic capacity management.

show abstract

Automatic Failure Diagnosis Support in Distributed Large-Scale Software Systems Based on Timing Behavior Anomaly Correlation

Cited by 37 publications

References 5 publications

Root cause detection in a service-oriented architecture

Root cause detection in a service-oriented architecture

Towards a Dependability Control Center for Large Software Landscapes (Short Paper)

Microservices for Scalability

Contact Info

Product

Resources

About