Topic

Exploration and Application Practices of Big Models in Observable AIOps

In the cloud-native era, Ops teams face significant challenges such as fluctuating data quality, insufficient labelling, and incomplete link information, obstacles that seriously affect the effective implementation of AIOps (intelligent operations). Uncertainty in monitoring data increases the complexity of O&M tasks, making it more difficult to extract useful information from massive data and respond quickly. The powerful emergent and inference capabilities of large models make it possible for AIOps to provide more accurate data correlation and diagnosis under complex architectures. The AliCloud Observable team explored the use of big models for knowledge query and PromQL query generation through AIOp combined with big model applications, which greatly simplified the process of querying and analysing monitoring data. In addition, the big model excels in providing context-explicit alerts, and through integration with ChatOps bots, a more natural and user-friendly interaction is achieved. These practices not only improve diagnostic efficiency, but also provide strong support for rapid resolution of system anomalies. However, large models, as a double-edged sword in AIOps implementations, also bring a new set of challenges, including model interpretability issues and integration with existing systems. The agent co-diagnostic solution based on the ReAct framework provides a possible solution path to these challenges, enabling AIOps to move beyond single automation tasks towards a more collaborative and adaptive approach. Going forward, we predict that big models will continue to play a key role in the observability space, driving further development and innovation in AIOps technologies. Technical Sharing Outline 1. Introduction: O&M Data Challenges in the Cloud Native Era ○ Analyse and monitor O&M data quality fluctuations, under-labelling and missing link information ○ Explore how these challenges affect the effective implementation of AIOps 2. AliCloud Observable Big Model Practice: Unlocking the Diagnostic Key for Complex Architectures ○ Sharing typical applications of big models in observable: knowledge quiz, PromQL generation ○ Explore the practice of big models in dealing with diagnostic problems of complex systems. ○ Discuss the use of big models to help ChatOps bots "speak human" in context-aware alerts. 3. The Double-Edged Sword of Big Models: Implications and Challenges for AIOps ○ Analyse how big models can bring changes to AIOps, and the new challenges they may bring. ○ Explore the value of agent collaborative diagnostic solutions based on the ReAct framework? 4. Summarise and look ahead: future exploration of observable big models ○ Provide an outlook on the future direction of big models to the observable domain