Articles

This article addresses the challenge of efficiently sizing MongoDB Atlas Search deployments, particularly for workloads characterized by large index sizes but moderate query and indexing rates. It highlights how previous node sizing options often led to costly overprovisioning of compute resources when only increased storage was required. To mitigate this, MongoDB introduces storage-optimized search nodes. The post provides a comprehensive breakdown of the core components influencing search node sizing, including methods for estimating data and index size, considerations for data ingestion and the indexing process, strategies for managing steady-state replication and lag, and techniques for optimizing query performance (QPS and latency). For each factor, practical considerations and optimization strategies are discussed. The new storage-optimized nodes are presented as a cost-effective alternative, offering significantly higher storage capacity (more than double) at a reduced price point. These nodes are ideal for storage-bound workloads and specific vector search scenarios where index size is the primary scaling factor. The article concludes by emphasizing the blend of art and science in effective deployment sizing and how these new nodes empower users to build more performant and cost-efficient Atlas Search solutions.

The manufacturing sector faces significant challenges, including costly downtime and labor shortages, making predictive maintenance a foundational pillar for operational excellence. This article introduces multi-agent AI systems as the next frontier for predictive maintenance, building upon generative AI and Retrieval-Augmented Generation (RAG). It presents a detailed blueprint for implementing such a system using MongoDB Atlas as the unified data layer, LangGraph for agent orchestration, and Amazon Bedrock for underlying foundational models. The proposed architecture employs a supervisor-agent pattern with specialized agents for failure analysis, work order generation, and optimal maintenance scheduling. The four-step process covers real-time anomaly detection using Atlas Stream Processing and Time Series data, automated root cause analysis with Atlas Vector Search, work order automation with a human-in-the-loop, and intelligent scheduling. The article highlights how MongoDB's capabilities enable scalable and responsive industrial AI solutions, offering a clear path to manufacturing excellence.