How HPE Zerto built an agentic troubleshooting system with Amazon Bedrock
aws ·
HPE Zerto has developed an agentic troubleshooting system powered by Amazon Bedrock to assist users in managing hybrid and multi-cloud infrastructures. This solution addresses the complexity of disaster recovery operations by providing a natural language interface that delivers actionable insights grounded in live environmental data. The architecture is deployed as an on-premises pod within the Zerto product, ensuring secure access to operational signals while sending only inference requests to AWS. An orchestrator agent manages specialized sub-agents for different components, such as the ZVM and VRA agents, which handle specific troubleshooting tasks using local tools and external knowledge bases. To ensure reliability, the team implemented a comprehensive evaluation framework using PyTest and the Strands Agents SDK to assess response quality, latency, and token usage. Security and governance are maintained through Amazon Bedrock Guardrails, which enforce content boundaries and compliance requirements before and after model inference. Since its release in the second quarter of 2026, over twenty percent of HPE Zerto customers have adopted this AI-assisted system, demonstrating strong demand for automated resilience management.
aws