Q: 2
A company has trained and deployed an ML model by using Amazon SageMaker. The company needs
to implement a solution to record and monitor all the API call events for the SageMaker endpoint.
The solution also must provide a notification when the number of API call events breaches a
threshold.
Use SageMaker Debugger to track the inferences and to report metrics. Create a custom rule to
provide a notification when the threshold is breached.
Which solution will meet these requirements?
Options
Discussion
Option D
Option D is the best fit. SageMaker automatically exports the Invocations metric to CloudWatch, so you just need to visualize and alarm on it. Debugger isn't meant for tracking endpoint API calls like this, pretty sure about D unless I'm missing something obvious.
C/D? CloudTrail might be useful for full API audit logs but for just tracking invocation counts and alerting, D is cleaner since Invocations metric is built in. I'm leaning D, anyone see a case for C here?
C vs D? The question asks about all API call events, but D is what I'd pick here.
A is wrong, D. The Invocations metric in CloudWatch tracks exactly what you need for SageMaker endpoints, and setting up an alarm is quick. Debugger isn't meant for tracking API call events directly. Think D fits best unless I'm missing something.
Had something really similar in a mock, picked D. Invocations metric gives you exactly the API call counts, and CloudWatch alarms let you alert right away when it hits the threshold. Way more straightforward than using CloudTrail or Debugger for this use case. Pretty sure that's what AWS expects here, but happy to hear other takes.
D imo. SageMaker pushes invocation metrics straight to CloudWatch, so adding the Invocations metric and setting an alarm is the cleanest option. No need for extra tools like CloudTrail here. Let me know if I'm missing something.
Kind of see the logic for C but still think D is the closest fit for monitoring invocation counts. Anyone picking B?
I was thinking C makes sense because CloudTrail captures all the API invocations, and you can use CloudWatch dashboards plus alarms for alerting. Might be overkill for just monitoring invocation counts, but should work. Not 100% though, maybe I'm missing a detail.
Probably D, Debugger doesn't track API events like invocations so A is misleading here.
Be respectful. No spam.
