Unattended Project Recommender aims to provide a unified tool to discover, reclaim, and shut down unattended cloud computing projects, using actionable and automatic machine learning-powered recommendations.
In enterprise environments, it is common for cloud resources to occasionally slip through the cracks. They can be difficult to identify and create problems for products teams in the future, including unwarranted waste.
What the survey found
Anodot conducted a survey recently which found that fewer than 20% of companies were able to immediately detect spikes in cloud costs. Additionally, 77% of companies had over $2 million in cloud costs and were caught off-guard when presented with the figures detailing how much they spent.
The tool is available through Google Cloud’s Recommender API and aims to identify abandoned projects, usually the result of billing, cloud service usage, network activity, and other signals.
Dima Melynk and Bakh Inamov, the product managers, say that the tool was first tested with teams at Google over this year, where it was used to clean up internal unattended projects and later, projects run by select Google Cloud customers.
How it works
Unattended Project Recommender analyzes usage activity across all projects in an organization, including service accounts, network ingress, and egress, API calls used, the number of active virtual machines (VMs), storage requests, services with billable usage, active project owners, and BigQuery jobs.
Based on various signals, the recommender can generate suggestions to clean up projects with low usage. ‘Low usage’ is defined using an ML algorithm that ranks the projects in an enterprise by the level of usage or makes recommendations to reclaim projects with high usage and no active project owners.