Based on the knowledge sources, there isn't specific information about recommended instance types or exact machine specifications for running Ory Kratos in production. However, I can provide some insights from the available information about Ory's scalability and deployment considerations. Ory services, including Kratos, are designed to scale efficiently. According to the documentation, Ory "scales effortlessly to thousands of pods without any additional work" and there's "no need for complex key-value stores or message queues to serve high traffic environments" [Scalability](https://www.ory.sh/docs/self-hosted/operations/scalability). For a real-world example, the documentation mentions that for sites with 1-1,000 daily active users, a setup might include: - Two virtual machines for failover - Each with 2 vCPUs and 4GB of RAM - Capable of running up to three Ory services - One small PostgreSQL instance with 100GB storage For sites with 1,000-20,000 daily active users, a larger setup is suggested: - Four machines with 4 vCPUs and 8GB RAM each - Two PostgreSQL instances with 4 vCPUs, 16GB RAM, and 500GB storage These specifications come from a cost comparison between self-hosting and using Ory Network [Ory Network or self-hosted?](https://www.ory.sh/blog/ory-network-or-self-hosting-explained-use-cases). One important consideration for Kratos specifically is the mail courier component, which should run as a singleton to avoid processing emails multiple times. In multi-instance deployments, this needs special consideration [Scalability](https://www.ory.sh/docs/self-hosted/operations/scalability). Also, be aware that password hashing with Argon2 can be resource-intensive and may require tuning to avoid performance issues or out-of-memory errors [Performance problems](https://www.ory.sh/docs/kratos/debug/performance-out-of-memory-password-hashing-argon2). For more detailed recommendations specific to your expected traffic and use case, you might want to reach out to Ory directly.