Google Cloud Vertex AI

이 페이지는 아직 한국어로 제공되지 않으며 번역 작업 중입니다. 번역에 관한 질문이나 의견이 있으시면 언제든지 저희에게 연락해 주십시오.

Overview

Google Cloud Vertex AI empowers machine learning developers, data scientists, and data engineers to take their projects from ideation to deployment, quickly and cost-effectively. Train high-quality custom machine learning models with minimal machine learning expertise and effort.

Setup

Installation

Metric collection

Google Cloud Vertex AI is included in the Google Cloud Platform integration package. If you haven’t already, set up the Google Cloud Platform integration first to begin collecting out-of-the-box metrics.

Configuration

To collect Vertex AI labels as tags, enable the Cloud Asset Viewer role.

You can use service account impersonation and automatic project discovery to integrate Datadog with Google Cloud.

This method enables you to monitor all projects visible to a service account by assigning IAM roles in the relevant projects. You can assign these roles to projects individually, or you can configure Datadog to monitor groups of projects by assigning these roles at the organization or folder level. Assigning roles in this way allows Datadog to automatically discover and monitor all projects in the given scope, including any new projects that may be added to the group in the future.

Log collection

Google Cloud Vertex AI logs are collected with Google Cloud Logging and sent to a Dataflow job through a Cloud Pub/Sub topic. If you haven’t already, set up logging with the Datadog Dataflow template.

Once this is done, export your Google Cloud Vertex AI logs from Google Cloud Logging to the Pub/Sub topic:

Go to the Google Cloud Logging page and filter Google Cloud Vertex AI logs.
Click Create Sink and name the sink accordingly.
Choose “Cloud Pub/Sub” as the destination and select the Pub/Sub topic that was created for that purpose. Note: The Pub/Sub topic can be located in a different project.
Click Create and wait for the confirmation message to show up.

Data Collected

Metrics

gcp.aiplatform.prediction.online.cpu.utilization (gauge)	Fraction of CPU allocated by the deployed model replica and currently in use. May exceed 100% if the machine type has multiple CPUs. Sampled every 60 seconds. After sampling data is not visible for up to 360 seconds. Shown as fraction
gcp.aiplatform.prediction.online.memory.bytes_used (gauge)	Amount of memory allocated by the deployed model replica and currently in use. Sampled every 60 seconds. After sampling data is not visible for up to 360 seconds. Shown as byte
gcp.aiplatform.prediction.online.prediction_latencies.samplecount (count)	Online prediction latency of the public deployed model. Sampled every 60 seconds. After sampling data is not visible for up to 360 seconds. Shown as microsecond
gcp.aiplatform.prediction.online.prediction_latencies.avg (gauge)	Average Online prediction latency of the deployed model. Shown as microsecond
gcp.aiplatform.prediction.online.prediction_count (count)	Number of online predictions. Shown as prediction
gcp.aiplatform.prediction.online.network.sent_bytes_count (count)	Number of bytes sent over the network by the deployed model replica. Sampled every 60 seconds. After sampling data is not visible for up to 360 seconds. Shown as byte
gcp.aiplatform.prediction.online.network.received_bytes_count (count)	Number of bytes received over the network by the deployed model replica. Sampled every 60 seconds. After sampling data is not visible for up to 360 seconds. Shown as byte
gcp.aiplatform.prediction.online.target_replicas (count)	Target number of active replicas needed for the deployed model. Sampled every 60 seconds. After sampling data is not visible for up to 120 seconds. Shown as worker
gcp.aiplatform.prediction.online.replicas (count)	Number of active replicas used by the deployed model. Sampled every 60 seconds. After sampling data is not visible for up to 120 seconds. Shown as worker
gcp.aiplatform.prediction.online.response_count (count)	Number of different online prediction response codes. Shown as response
gcp.aiplatform.prediction.online.error_count (count)	Number of online prediction errors. Shown as error
gcp.aiplatform.online_prediction_requests_per_base_model (count)	Online prediction requests per minute per project per base model. Shown as request
gcp.aiplatform.prediction.online.accelerator.duty_cycle (gauge)	Fraction of CPU allocated by the deployed model replica and currently in use. May exceed 100% if the machine type has multiple CPUs. Sampled every 60 seconds. After sampling data is not visible for up to 360 seconds. Shown as fraction
gcp.aiplatform.prediction.online.accelerator.memory.bytes_used (gauge)	Amount of accelerator memory allocated by the deployed model replica. Shown as byte
gcp.aiplatform.prediction.online.private.prediction_latencies.avg (gauge)	Average Online prediction latency of the private deployed model. Shown as microsecond
gcp.aiplatform.prediction.online.private.prediction_latencies.samplecount (count)	Online prediction latency of the private deployed model. Sampled every 60 seconds. After sampling data is not visible for up to 360 seconds. Shown as microsecond
gcp.aiplatform.prediction.online.private.response_count (count)	Online prediction response count of the private deployed model. Shown as response
gcp.aiplatform.quota.online_prediction_requests_per_base_model.exceeded (count)	Number of attempts to exceed the limit on quota metric aiplatform.googleapis.com/onlinepredictionrequestsperbase_model. Shown as error
gcp.aiplatform.quota.online_prediction_requests_per_base_model.limit (gauge)	Current limit on quota metric aiplatform.googleapis.com/onlinepredictionrequestsperbase_model. Shown as request
gcp.aiplatform.quota.online_prediction_requests_per_base_model.usage (count)	Current usage on quota metric aiplatform.googleapis.com/onlinepredictionrequestsperbase_model. Shown as request