The cloud market is undergoing fundamental changes. Traditional infrastructure models, designed for human traffic, are no longer coping with the realities of the artificial intelligence era. In response to the challenges of this new era, AWS has introduced an updated version of OpenSearch Serverless — a fully managed database created specifically for handling agent workloads.
The Problem: From Experiments to Production Reality
Tia White, General Manager of Amazon OpenSearch Service, notes that AI agents have ceased to be just experimental projects and have moved to the stage of industrial implementation. This change has radically transformed traffic models. Agents are capable of generating instantaneous spikes in activity, creating numerous sub-tasks that access hundreds of databases, search for documents, and call APIs within seconds, only to stop just as quickly.
Previous infrastructure was not designed for such spikes. Traditionally, working with OpenSearch required allocating EC2 instances, manually configuring clusters, managing scaling, as well as handling updates and backups. In the new reality, such approaches become inefficient and expensive.
The Solution: Architecture Separating Compute from Storage
The key innovation of the next-generation OpenSearch Serverless lies in the separation of compute resources and storage. This allows the service to scale compute power in seconds to handle peak loads and scale down to zero during idle times.
Previously, in the earlier version of Serverless, storage and compute were linked. Clients had to maintain at least one running instance, which meant reserved compute resources were present even if they were not being used. The new architecture eliminates this issue.
Now, clients pay only for the resources consumed by active workloads. The service uses OpenSearch Compute Units (OCU) for dynamic scaling and billing. This ensures significant savings: AWS claims a cost reduction of up to 60% compared to the cost of OpenSearch Service clusters provisioned for peak load.
Integration with the Ecosystem and Global Trends
At launch, OpenSearch Serverless will be integrated with leading AI development platforms, such as Vercel and Kiro. This will allow developers to deploy production-ready backends for search and vectorization without the need to manage complex infrastructure.
The transformation affects not only AWS. The expanded use of agent AI is changing the entire cloud industry. Microsoft has released updates for the Azure platform aimed at provisioning resources for AI agent activity spikes and memory sharing. Cloudflare has also introduced infrastructure to provide agents with persistent environments and instant scalability.
The scale of changes is confirmed by statistics. According to Cloudflare, over the last six months, bots accounted for nearly a third (31%) of all HTTP traffic. Approximately a quarter of all bot requests were generated by search crawlers, search engines, and voice assistants.
Lay I Olsen, Senior Product Manager at Cloudflare, predicts that non-human traffic will exceed human traffic around the first half of 2027. The updates from AWS are a direct response to this inevitable paradigm shift in the digital world.