Pinecone has introduced Dedicated Read Nodes (DRN) in public preview to address the needs of applications requiring consistent high throughput and low latency, such as billion-vector semantic searches and real-time recommendation feeds. Unlike their On-Demand service, which is ideal for bursty workloads, DRN offers reserved capacity, ensuring predictable performance and cost by using dedicated, provisioned nodes, which isolate performance and maintain data in memory and local SSDs to avoid latency from cold fetches. Customers can choose between DRN and On-Demand services based on workload requirements, allowing for scalability and efficient cost management without compromising performance. DRN supports scaling through replicas, which increase throughput and resilience, and shards, which expand storage capacity, offering a flexible solution for enterprise-grade vector database applications.