Temporal Scalability

Question

I am looking to use temporal to stitch my customers journey in an e-commerce platform. The workflow is supposed to start when a user clicks on 'search', and supposed to end when the customer finally places the order. Current objective of this workflow is to track user's journey in the e-commerce funnel, and provide experience to the dropped off customers for resuming there transaction on the platform

Temporal seems like a really good fit with the offerings around resilience, very expressive state-machine support (Ex : scheduled transitions based on timers, etc), decoupling business logic vs orchestration, durable execution, etc.

Though looking very lucrative from the POC's/reading i have done, looking for suggestions around if temporal has been battle tested at the scale of

5000 new workflows initiated every second (search scale)
Each workflow spanning max for 30 days
Each workflow has maximum 4 state transitions.

Any insights are welcome. Thanks

Tom Wheeler · Accepted Answer · 2025-06-18 14:34:25Z

There are definitely Temporal users running at a scale that exceeds any of the individual constraints you describe. However, it seems to me that if you're talking about all of them together, you'd eventually reach a capacity problem with any system because you're starting new Workflow Executions about 2.5 million times faster than the 30 days you say it would take for them to finish.

In other words, if you're spawning new Workflow Executions at the rate of 5,000 per second and they have a lifetime of 30 days, then after 2 seconds you'd have 10,000, after 3 seconds you'd have 15,000, and so on. At the end of the first day, you'd have 432,000,000 and that would keep increasing until the first one finally completed 30 days later (by which point you'd have 12,959,999,999 Workflow Executions running).

Collectives™ on Stack Overflow

Temporal Scalability

1 Answer 1

Comments

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Related