Predictive Server Log Management
3AI September 6, 2020
Leading USA manufacturing company
Problem Statement
- Leading USA manufacturing company has multiple IT Infrastructure requests to handle. Some of these requests deal with application performance issues and a few critical ones are downtime/crash details.
- The team shared a sample set from a particular line of business with us, we studied this to bring in a preemptive solution for effective application monitoring.
Sample Event Log File:
Analytics Led Approach
- A predictive model was developed to predict “time-to-event” for various outage issues. It involved advanced techniques like text mining, association rules and random forests using R as the coding language
- Not only an estimation of which event will occur (e.g. Server Down instance) we were also able to predict in which time window this will happen next.
- Additionally, this was dynamic, as we also predicted which new events will re-trigger the time for ‘Server Down’.
Critical Success Factors
- Focused Targeting: Based on model, network team could target and intervene issues that were likely to occur in near future.
- Analytics helps in taking proactive steps to reduce unexpected outages.
- There would be both hard and soft cost savings due to this model. Direct hours saved, that would be otherwise spent by network admin in fixing issues, this helps them to pre-empt the problem in first place.
