Outlier Analysis in Live Systems from Application Logs: Unterschied zwischen den Versionen

Version vom 28. Mai 2021, 09:03 Uhr

Vortragende(r)	Wenrui Zhou
Vortragstyp	Proposal
Betreuer(in)	Edouard Fouché
Termin	Fr 11. Juni 2021
Vortragsmodus
Kurzfassung	Modern computer applications tend to generate massive amounts logs and have become so complex that it often is difficult to explain why a specific application has failed. In this work we want detect and explain such failure by detecting outliers from application logs. This is challenging because (1)Log is unstructured text streaming data. (2)labelling application log is labor intensive and inefficient. Logs are similar to natural languages. Recent advances in deep learning have shown great performance in Natural Language Processing (NLP) tasks. Based on these, we investigate how state-of-the-art sequence-to-sequence frameworks with attention mechanisms can detect and expolain outliers from applications logs. We plan to compare our framework against state-of-the-art log outlier detectors, based on existing outlier detection benchmarks.

@@ Zeile 5: / Zeile 5: @@
 |betreuer=Edouard Fouché
 |termin=Institutsseminar/2021-06-11
-|kurzfassung=Outlier Detection in today’s application logs is a difficult task because such applications generate massive amounts of unstructured logs, and the formate of such logs differs from one application to another. Since logs are similar to natural languages and state-of-the-art deep learning algorithms have achieved fantastic performance in natural language processing, we utilize state-of-the-art seq2seq frameworks and their attention mechanisms to detect and explain outliers in application logs. We test our framework with several outlier detection benchmarks and achieve comparable performance to state-of-the-art log outlier detection frameworks.
+|kurzfassung=Modern computer applications tend to generate massive amounts logs and have become so complex that it often is difficult to explain why a specific application has failed. In this work we want detect and explain such failure by detecting outliers from application logs. This is challenging because (1)Log is unstructured text streaming data. (2)labelling application log is labor intensive and inefficient.
+Logs are similar to natural languages. Recent advances in deep learning have shown great performance in Natural Language Processing (NLP) tasks. Based on these, we investigate how state-of-the-art sequence-to-sequence frameworks with attention mechanisms can detect and expolain outliers from applications logs. We plan to compare our framework against state-of-the-art log outlier detectors, based on existing outlier detection benchmarks.
 }}