Does EventBridge Scheduler Flexible Time Windows work?

Flexible time window helps in making sure that none of the event rule gets missed if there is surge in rule triggers at same point of time, AWS gets the flexibility to distribute that surge and would make sure that all the rules get triggered within their flexible time window.

With flexible time window EventBridge Scheduler invokes the task within that timeframe. This setting would help to distribute the invocations across time and manage the downstream service limits.

What is happening in your case is absolutely fine and as expected.

When you configure your schedule with a flexible time window, it invokes the target within the time window you set. For example, if you configure a 15 minute flexible time window for a schedule that runs every hour, it invokes the target within 15 minutes after the scheduled time.

So, in your case with 20 minutes of frequency, let's say next schedule is 7:40PM and you have flexible time window set as 15 minutes, then your rule would trigger the target no later than 7:55PM.

You may see variance if there is surge/spike in event rule trigger that have same scheduled time, as AWS would get flexibility to balance the load and makes sure that event rule would be triggered with the Schedule+flexible time window time. But with flexible time window you may not see variance necessarily, as there may not be enough load and AWS has bandwidth to invoke those rules on the exact schedule.

Reference:

Configuring flexible time windows

Hope this explanation helps.

Comment here if you have additional questions, happy to help.

Abhishek

While what you say is plausible, this is not now Amazon is advertising what this feature actually does.

For example on their blog introducing EventBridge Scheduler they have this to say about Time Windows:

Time window allows you to start a schedule within a window of time. This means that the scheduled tasks are dispersed across the time window to reduce the impact of multiple requests on downstream services.

Disperse is to spread or distribute across the time window (presumably to mitigate the thundering herd problem). It says not one iota about AWS surge or load. Infact none of the documentation talks about AWS spreading the execution over a time window due to load, and shouldn't this be the default action (to execute at the earliest possible time after the scheduled time) anyway if there is a surge in load?

As best as I can tell, the flexible time window only applies to the first invocation. That is, for your example of a schedule that's a rate of 20 minutes and a flexible window of 5 minutes, it will pick a random time within the first 5 minutes after enabling the schedule to run the first time. But then after that, it will continue every 20 minutes exactly from that first run.

I think the idea is that if one had multiple schedules set up that way and enabled them all at once, they would each pick random times but then comply with the specified rate.

Even for cron expressions, I haven't tested it much but it looks like it will similarly just calculate a random offset once, and then just add that same amount to each individual invocation.