Minimize Cold Starts with Recommender
Recommender is a service that automatically provides recommendations and insights for using resources on Google Cloud, based on heuristic methods, machine learning, and current resource usage. Each recommendation includes a link you can click to put the recommendation into effect for your function.
This topic shows how to use Recommender to minimize cold starts:
- Detect functions that could benefit from setting up the minimum instances (in other words, functions that are likely to be negatively affected by cold start issues).
- Inform users about the opportunity to eliminate cold start issues.
- Let users set up the minimum instances directly from the recommendation system.
Recommendation availability after deployment
Recommender automatically provides recommendations for a function as soon as a week after it has been deployed. After this period of time, recommendations for the function are displayed with the function on the function list page in the Cloud Run functions Overview page in the Google Cloud console and in the Recommendation Hub.
Alternate ways of using recommendations
In addition to the use of recommendations covered on this page inside the Cloud Run functions UI, recommendations are also available through the following:
Viewing and accepting recommendations for Cloud Run functions
To view and accept a recommendation in the Cloud Run functions user interface:
Look for functions in the list that have a Performance light bulb icon in the Recommendation column.
Click Performance to view details for the recommendation:
- An overview of why this recommendation is being suggested.
- The recommendation itself.
If you decide to apply the recommendation, set the number of minimum instances and click Apply.
This redeploys the function with the new minimum instance value. After this, the recommendation will no longer be displayed next to the function.
Dismissing a recommendation
Click Dismiss if you want to dismiss the recommendation without applying it. This prevents the recommendation for that function from appearing again for 30 days.
Viewing recommendations in Recommendation Hub
You can also view recommendations in the Recommendation Hub.
For more information refer to the Recommendation Hub
Getting started
documentation.