Improved log anomaly detection leads to faster time to value.
The sooner an artificial intelligence (AI) model can be ready to use, the better mean-time-to-detect we can achieve. Pretty straightforward, right? Faster detection is the crucial first step toward faster diagnosis and resolution. In a particular scenario with a client, IBM Cloud Pak® for Watson AIOps was able to identify log anomalies, reducing the mean-time-to-detect (MTTD) and mean-time-to-identify (MTTI) from 60 hours to ZERO.
In the world of AI for IT Operations (AIOps), context is king. You can’t understand an issue while missing parts of the narrative. It’s like trying to complete a puzzle without all the pieces. You need to gather all the metrics, logs, changes, tickets, topology, alerts and events to stitch together the story of an incident. With IBM Cloud Pak for Watson AIOps, that’s what we do. We harness the power of real AI to learn the patterns of your data, extract the major points of interest and shape a congruent story of a problem:
Logs are one of the key parts of an incident’s story, but they are arguably one of the most difficult pieces of data to analyze. Although logs contain important context information — timestamps, errors, warnings, etc. — they’re semi-structured in nature, extremely voluminous, varied in format and may contain mixed languages. All of these factors make it quite difficult to extract the information you need. It’s not easy, especially in real-time.
Luckily with IBM Cloud Pak for Watson AIOps, we utilize state-of-the-art and multi-patented log anomaly detection technology that is capable of automatically parsing IT application and infrastructure logs. IBM Watson automatically learns the normal log patterns from training data, understands their semantic meaning and detects anomalies in real-time, using deep learning algorithms:
3.2 release of IBM Cloud Pak for Watson AIOps
Pretty cool stuff to have under your belt when fighting off IT incidents, but what’s the use of investing in technology when you can’t quickly see the return in value? That is usually the roadblock when talking AIOps. They don’t call it machine learning for nothing. AI needs time to learn. But how much time? In our 3.2 release of IBM Cloud Pak for Watson AIOps, our team has reduced the training time down to 30 minutes, for 1 million logs. That’s it. This is down from two weeks in our previous release!
A vital part of the AIOps mission is delivering value as immediately as possible. This isn’t supposed to be easy, but IBM is proving that it can handle the burden. This way, you can keep availability at an all-time high, while driving innovation continuously in your business.
Ready to learn more? Check out my webinar and learn more about what you can expect from the IBM Cloud Pak for Watson AIOps 3.2 release.