Shoutout to Surya Nimmagadda and Alex Lau for their collaboration on this blog.
Next Generation AIOps
Two simultaneous and dramatic shifts combine in next generation AIOps: ease of use and data-centric learning:
- Data-centric learning eliminates AIOps solution configuration and connects the dots within and across different data types.
- Collaboration-centric operations team members get their work done anywhere, from any device, at any time, natively within the environment they use everyday.
This blog focuses on data-centric learning while a future blog will focus on collaboration.
The Selector AI AI/ML analytics engine has four foundational differentiators:
- Zero touch operation
- Any data analysis
- Dynamic thresholding
- Root cause ranking
Zero touch operation
Early attempts at applying AI/ML to network operations were hampered by complex, time consuming setup and maintenance. At best, this resulted in a lengthy period of time before any value was realized. Often the result was worse, customers giving up on a tool. The promise of AI/ML was not realized.
Selector AI is a next generation, operations-centric, AIOps solution, designed specifically to address issues in previous generations. The power of AI/ML in a fully automated, no-config solution, with powerful collaboration integration.
Some network operations teams have a large number of people and the ability to invest heavily in new technologies. Most operations teams do not have the necessary AI/ML investment capacity or skillset. For these teams, a simple to use solution with common integrations is essential. Network operations teams need not be exposed to the intricacies of data ingestion, data normalization, AI/ML workflow construction, AI/ML algorithms, or data query. Operation must be simple, intuitive, and powerful.
Any Data Analysis
Many tools are focused on numerical data, for example metric measurement. Sometimes tools are also focused on a specific aspect of networking, for example traffic analysis. While these approaches have had their uses, they fail to leverage all the data that is available, to create holistic, easy to understand, and actionable insights.
One characteristic of a next generation AIOps solution is the ability to rapidly identify root cause across multiple data types. For example, Selector AI already learns from logs, events, configuration and more. Not just ingesting data and reformatting it for visual display, but normalizing, augmenting, correlating, clustering, and filtering. Root cause ranked and easy to see because the noise is filtered out. Goodbye monitoring, hello observability and actionable insight.
Dynamic thresholding
Thresholds are among the biggest problems for network operations teams. Operations teams are faced with two difficult alternatives: Accept ineffective heuristics for thresholds and be overwhelmed with false alerts or hand-craft thresholds for an exploding number of endpoints. There is an important third alternative delivered by Selector AI: dynamic thresholding.
Instead of configuring and constantly tuning millions of thresholds to the precise level required to be effective, dynamic thresholding learns normal. Anomalies then become statistical variants from an established baseline. Importantly, Selector AI can scale this approach to many millions of thresholds, a necessity in today’s virtual, overlay, and IoT networks.
Root Cause Ranking
Selector AI’s automated ranking algorithm rapidly draws operations attention to which resources are most likely diverging from normal and are most often indicated as being a participant in a network issue. Both of these approaches automate the sifting of enormous amounts of data in addition to connecting the dots within and across data types. The Selector AI expert analysis tool also provides timelines and analysis of how different data types are associated, for example, where a configuration change is on a timeline, and whether problems started to arise as a result of that change.
Conclusion
Next generation AIOps from Selector AI, is fundamentally about a paradigm shifting change in experience and analytics. One critical piece of experience is zero touch operations: end to end automation of data collection, analysis, and root cause ranking. The core modeling is consistent regardless of the data type, so only data declarations need to change, and only those declarations need to occur if the data source is not already pre-integrated. No workflow construction for the solution is required. No normalization, filtering, correlation, or clustering code has to be written. Simple, automated, and insightful. A new generation of analytics without the learning curve. You can still be a data scientist, but you do not have to be. Let Selector AI be your data whisperer and dramatically reduce the overhead of setting up thresholds, ingesting data, connecting the dots, and ranking root causes.