AIOps, or artificial intelligence for IT operations, uses AI and ML technologies alongside big data, data integration and automation to help make IT operations smarter and more predictive. AIOps has come around as a response to a pressing need for optimizing operations and minimizing risks to the IT infrastructure in the modern IT ecosystem.
Gartner’s Distinction of Domain-centric and Domain-agnostic AIOps
Gartner defined two high-level categories of AIOps as domain-centric and domain-agnostic. Domain-centric AIOps platforms focus on homogenous, first-party datasets and offer AI capabilities to solve specific business use cases, such as network and application diagnostics.
However, domain-agnostic AIOps tools combine diverse datasets and data formats and integrate the analyses into actionable insights. Let’s delve deeper into each before discovering a secret third option for the future-proof organization.
What are Domain-centric AIOps tools and their use cases?
Domain-centric AIOps tools are usually applicable to a company’s singular distinct network or endpoint, so the range of their impact is limited and narrow. These AIOps tools come in handy when an organization wants to focus solely on a domain and benefit from AI and automation applied to the data from that domain. However, if an organization wants to implement an AIOps solution across its IT environment, a domain-centric AIOps tool wouldn’t suffice.
For instance, an IT monitoring-specific tool is a domain-centric AIOps tool if it applies AI/ML algorithms to the localized area of IT monitoring. Another typical use case of a domain-centric AIOps solution is an Application Monitoring tool or APM that measures application latency. The APM tool identifies anomalies in application performance and triggers an action when the latency is higher than a threshold using AI/ML.
Vendors who originally focused on incident management and later applied AI/ML to incident-specific use cases also fall under domain-centric AIOps vendors. Domain-centric tools relieve a specific pain in a specific part of an organization.
They are relevant to organizations with limited data variety, who prioritize a small number of use cases and have little need to integrate and process data across silos.
What are Domain-agnostic AIOps tools and their use cases?
Domain-agnostic AIOps tools access data from various sources in different formats, including current and historical data. These solutions then process this data and optimize a company’s IT infrastructure for performance and resilience.
Domain-agnostic AIOps tools gather data from applications, microservices, incidents, cloud-based systems, and from any other observability and infrastructure monitoring tools. A domain-agnostic approach is to then normalize the alerts into a data model, enrich them with context and correlate events together into a logical and cohesive incident that can be acted upon.
Domain-agnostic AIOps tools don’t just focus on alleviating a specific pain but on bringing strategic outcomes for an organization. If a company wants to make IT operations more innovation-friendly, seamless and efficient, they may want to consider a domain-agnostic AIOps tool.
These tools introduce flexibility in processing diverse datasets across silos and offer a progressive roadmap. They enable a host of use cases across IT ops management, such as DevOps, SRE, I&O, cybersecurity and compliance. Domain-agnostic AIOps tools go beyond anomaly detection to track user behavior, identify underlying business opportunities and boost customer engagement.
What is Data-centric AIOps and its use cases?
Most organizations still choose between a domain-centric and domain-agnostic AIOps approach. A third approach is quickly becoming popular and mission-critical. Data-centric artificial intelligence relies on data scientists to define an entire pipeline from data cleansing to ingestion through model training, eliminating the need for knowledge of AI algorithms.
Data-centric AI says, “let’s start with clean data and train an algorithm on it instead of training an algorithm first and then cleansing a messy dataset.” Data-centric AI fills gaps in an organization through augmentation, interpolation and extrapolation.
Traditional ETL or ELT techniques can no longer be used for data preparation as data now comes from distributed and hybrid applications, where the MELTs are steaming, events-based and alerts-based.
AIOps then needs to process all data types to improve data quality, which is only achievable with real-time observability pipelines. Data-centric AIOps can perform the following functions to unlock potential across an organization:
- Data integration – Data-centric AIOps leverages low-code and no-code bots to integrate monitoring, observability and APM sources, pulling insights across all.
- Data ingestion, routing and compliance – The streaming data is then normalized, duplicated, redacted and ingested into the AIOps platform over low-latency data fabric for the core, edge and multi-cloud applications. It retains a full-fidelity copy in low-cost archival for compliance purposes while routing other processed copies to concerned stakeholders.
- Enrichment and contextualization – Automated pipelines discover topology in real-time and attributes from element managers, creating a context for correlation. External feeds can also be utilized to enrich the data.
- Correlation and continuous AI/ML – Many correlation pipelines reduce and suppress noise to perform root cause analysis using an incident recommendation engine. The data-centric AIOps tool uses all data types, alerts, metrics and traces to auto-remediate issues, create incidents, prioritize actions and route recommendations to concerned stakeholders.
- Anomaly detection – Anomaly detection regression pipelines build logs for dynamic baselining, anomaly detection, forecasts and predictions.
- Observability dashboards – Self-service, dynamic dashboards offer insights into business value and economic impact across observability, AIOps and automation domains.
The Criticality of Data-centric AIOps for Modern IT Monitoring
Here’s why data-centric AI is critical for AIOps:
- ML algorithms require a lot of data to achieve accuracy in predictions and actionability of insights.
- AIOps functions constantly evolve, requiring to incorporate data into AIOps constantly to improve performance.
- Data-centric AI helps identify patterns and relationships that are otherwise hard to discern automatically.
- Data-centric AI can identify outlying patterns in behavior that can subsequently be used for further investigation and action.
- Automating the collection and analysis of data can improve AIOps efficiency through data-centric AI.
An RDAF platform can help actualize data-centric AIOps in your organization. A pre-requisite to using a data-centric AIOps tool can be to automate data streams, use no-code and low-code pipelines and perform in-place analytics across disparate data sources, irrespective of where they reside.