Stateless SF app VS Stateful SF app

Question

I want to compare the "recovery" time in a node down scenario for stateful/stateless apps.

I think for a stateful app, it needs to promote a secondary replicate to primary.

For stateless app, it requires to start a new partition in a different node.

I suppose stateful app should be faster, but how fast it can be compared with stateless app? Is there any range I can refer to?

Answer 1

Hi Yicong Wu,

Welcome to the Microsoft Q&A Platform! Thank you for asking your question here.

The recovery time in a node-down scenario depends on multiple factors, including the application type (stateful vs. stateless), the underlying infrastructure, architecture, infrastructure, and specific implementations.

Stateful App Recovery Time: The recovery time depends on the consistency model (e.g., strong consistency vs. eventual consistency) and the size of the data being managed. For well-optimized systems (e.g., distributed databases like Cassandra, MongoDB, or Kafka), recovery can happen in seconds to a few minutes. If the system requires extensive data synchronization or has complex failover mechanisms, recovery can take longer.

Stateless App Recovery Time: Stateless applications do not need to manage data replication or synchronization, which simplifies recovery. For containerized applications (e.g., Kubernetes pods), recovery can be very fast, often in seconds and with highly optimized orchestration systems, recovery can be almost instantaneous.

In summary, while stateful applications provide continuity by retaining session information, their recovery in node-down scenarios can be more complex and time-consuming. Stateless applications, by not retaining session state, often achieve faster recovery times, enhancing fault tolerance and resilience.

If you have any further queries, please do let us know. If the answer is helpful, please click "Accept Answer" and "Upvote it" as it can be helpful to others in the community.