Monitoring Amazon SageMaker with Amazon CloudWatch
You can monitor Amazon SageMaker using Amazon CloudWatch, which collects raw data and processes it into readable, near real-time metrics. These statistics are kept for 15 months, so that you can access historical information and gain a better perspective on how your web application or service is performing. However, the Amazon CloudWatch console limits the search to metrics that were updated in the last 2 weeks. This limitation ensures that the most current jobs are shown in your namespace. To graph metrics without using a search, specify its exact name in the source view. You can also set alarms that watch for certain thresholds, and send notifications or take actions when those thresholds are met. For more information, see the Amazon CloudWatch User Guide.
Amazon SageMaker model training jobs and endpoints write CloudWatch metrics and logs. The following tables list the metrics and dimensions for Amazon SageMaker.
Endpoint Invocation Metrics
The AWS/SageMaker namespace includes the following request metrics from calls
to InvokeEndpoint
.
Metrics are available at a 1-minute frequency.
| Metric | Description |
|---|---|
Invocation4XXErrors |
The number of Units: None Valid statistics: Average, Sum |
Invocation5XXErrors |
The number of Units: None Valid statistics: Average, Sum |
Invocations |
The To get the total number of requests sent to a model endpoint, use the Sum statistic. Units: None Valid statistics: Sum, Sample Count |
InvocationsPerInstance |
The number of invocations sent to a model, normalized by Units: None Valid statistics: Sum |
ModelLatency |
The interval of time taken by a model to respond as viewed from Amazon SageMaker. This interval includes the local communication times taken to send the request and to fetch the response from the container of a model and the time taken to complete the inference in the container. Units: Microseconds Valid statistics: Average, Sum, Min, Max, Sample Count |
OverheadLatency |
The interval of time added to the time taken to respond to a client request by Amazon
SageMaker
overheads. This interval is measured from the time Amazon SageMaker receives the request
until it
returns a response to the client, minus the Units: Microseconds Valid statistics: Average, Sum, Min, Max, Sample Count |
Dimensions for Endpoint Invocation Metrics
| Dimension | Description |
|---|---|
EndpointName, VariantName |
Filters endpoint invocation metrics for a |
Training Job, Batch Transform Job, and Endpoint Instance Metrics
The /aws/sagemaker/TrainingJobs, /aws/sagemaker/TransformJobs and
/aws/sagemaker/Endpoints namespaces include the following metrics for the training
jobs and endpoint instances.
Metrics are available at a 1-minute frequency.
| Metric | Description |
|---|---|
CPUUtilization |
The percentage of CPU units that are used by the containers on an instance. The value
can range between 0 and 100, and is multiplied by the number of CPUs. For example,
if there
are four CPUs, For training jobs, the value is the CPU utilization of the algorithm container on the instance. For batch transform jobs, the value is the CPU utilization of the transform container on the instance. For endpoint variants, the value is the sum of the CPU utilization of the primary and supplementary containers on the instance. Units: Percent |
MemoryUtilizaton |
The percentage of memory that is used by the containers on an instance. This value can range between 0% and 100%. For training jobs, the value is the memory utilization of the algorithm container on the instance. For batch transform jobs, the value is the memory utilization of the transform container on the instance. For endpoint variants, the value is the sum of the memory utilization of the primary and supplementary containers on the instance. Units: Percent |
GPUUtilization |
The percentage of GPU units that are used by the containers on an instance. The value
can range between 0 and 100 and is multiplied by the number of GPUs. For example,
if there
are four GPUs, For training jobs, the value is the GPU utilization of the algorithm container on the instance. For batch transform jobs, the value is the GPU utilization of the transform container on the instance. For endpoint variants, the value is the sum of the GPU utilization of the primary and supplementary containers on the instance. Units: Percent |
GPUMemoryUtilization |
The percentage of GPU memory used by the containers on an instance. The value can
range
between 0 and 100 and is multiplied by the number of GPUs. For example, if there are
four
GPUs, For training jobs, the value is the GPU memory utilization of the algorithm container on the instance. For batch transform jobs, the value is the GPU memory utilization of the transform container on the instance. For endpoint variants, the value is the sum of the GPU memory utilization of the primary and supplementary containers on the instance. Units: Percent |
DiskUtilization |
The percentage of disk space used by the containers on an instance uses. This value can range between 0% and 100%. This metric is not supported for batch transform jobs. For training jobs, the value is the disk space utilization of the algorithm container on the instance. For endpoint variants, the value is the sum of the disk space utilization of the primary and supplementary containers on the instance. Units: Percent |
Dimensions for Training Job, Batch Transform Job, and Endpoint Instance Metrics
| Dimension | Description |
|---|---|
Host |
For training jobs, the value for this dimension has the format
For batch transform jobs, the value for this dimension has the format
For endpoints, the value for this dimension has the format |
