Articles in Analytics

Customers using Snowflake's Workload Identity Federation (WIF) with AWS Glue may encounter a "No AWS region was found" error because the Glue proxy layer blocks boto3's automatic region detection. Thi...

This article provides a decision framework for AWS customers running Amazon OpenSearch Service indices exceeding 500 GB per shard who question whether reindexing is necessary when search latency is ac...

After setting up a zero-ETL integration between DynamoDB and Amazon Redshift, some tables sync successfully while others repeatedly fall into a "Failed" state. This article explains how to systematica...

When you resize an Amazon Redshift cluster (changing the node type or number of nodes), you can validate the performance impact by capturing key metrics before and after the operation. This article pr...

Use Athena CTAS and Iceberg time travel to recover a writable table from a read-only S3 Tables replica

This article provides a detailed, step-by-step guide for AWS customers who have very large indices (>1 TB) in their OpenSearch cluster and want to reduce shard size for better performance and stabilit...

Swap activity on Amazon RDS for PostgreSQL is one of the most frequently misunderstood signals during memory-related investigations. Many users see SwapUsage > 0 in CloudWatch and assume something is ...

This article explains how to connect Amazon Quick with external tools and systems using Model Context Protocol (MCP). Many organizations want their AI agents to not only answer questions but also tak...

Amazon OpenSearch Service now supports OpenSearch 3.x, which introduces significant performance improvements through Apache Lucene 10, JVM 21, and enhanced vector search capabilities. Upgrading from 2...

Running Spark on EMR with KMS-encrypted S3 data? Every object read triggers a kms:Decrypt API call — and at scale, those costs add up fast. If your compliance requirements prevent switching to S3 Buck...

A field guide for PyTorch, TensorFlow, Spark, and Kubernetes workloads reading training data from Amazon S3 Express One Zone directory buckets.

This article explains when the MSK Connect service execution role is actually used, why worker log delivery does not use this role, and how to accurately determine whether the role is in use.