Los productos Google Cloud TPU ponen las ventajas de las Tensor Processing Units (TPU) a disposición de todos los investigadores de ML, ingenieros de ML, desarrolladores y científicos de datos que ejecutan modelos de ML de última generación, a través de un recurso informático en la nube escalable y fácil de utilizar.
Utiliza la integración de Google Cloud Platform con Datadog para recopilar métricas de Google Cloud TPU.
Configuración
Instalación
Para utilizar Google Cloud TPU, sólo necesitas configurar la integración Google Cloud Platform.
Recopilación de logs
Los logs de integración se recopilan con Google Cloud Logging y se envían a una tarea de Dataflow a través de un tema Cloud Pub/Sub. Si aún no lo has hecho, configura la generación de logs con la plantilla Dataflow de Datadog.
Una vez hecho esto, exporta tus logs de Google Cloud TPU de Google Cloud Logging al tema Pub/Sub:
- Ve a la página de Google Cloud Logging y filtra logs de Google Cloud TPU.
- Haz clic en Create Export (Crear exportación) y asigna un nombre al sumidero.
- Elige “Cloud Pub/Sub” como destino y selecciona el tema Pub/Sub creado para tal fin. Nota: El tema Pub/Sub puede estar ubicado en un proyecto diferente.
- Haz clic en Crear y espera a que aparezca el mensaje de confirmación.
Recopilación de datos
Métricas
gcp.tpu.cpu.utilization
(gauge) | Utilization of CPUs on the TPU Worker as a percent.
Shown as percent |
gcp.tpu.memory.usage
(gauge) | Memory usage in bytes.
Shown as byte |
gcp.tpu.network.received_bytes_count
(count) | Cumulative bytes of data this server has received over the network.
Shown as byte |
gcp.tpu.network.sent_bytes_count
(count) | Cumulative bytes of data this server has sent over the network.
Shown as byte |
gcp.tpu.accelerator.duty_cycle
(count) | Percentage of time over the sample period during which the accelerator was actively processing
Shown as percent |
gcp.tpu.instance.uptime_total
(count) | Elapsed time since the VM was started, in seconds.
Shown as second |
gcp.gke.node.accelerator.tensorcore_utilization
(count) | Current percentage of the Tensorcore that is utilized.
Shown as percent |
gcp.gke.node.accelerator.duty_cycle
(count) | Percent of time over the past sample period (10s) during which the accelerator was actively processing.
Shown as percent |
gcp.gke.node.accelerator.memory_used
(count) | Total accelerator memory allocated in bytes.
Shown as byte |
gcp.gke.node.accelerator.memory_total
(count) | Total accelerator memory in bytes.
Shown as byte |
gcp.gke.node.accelerator.memory_bandwidth_utilization
(count) | Current percentage of the accelerator memory bandwidth that is being used.
Shown as percent |
gcp.gke.container.accelerator.tensorcore_utilization
(count) | Current percentage of the Tensorcore that is utilized.
Shown as percent |
gcp.gke.container.accelerator.duty_cycle
(count) | Percent of time over the past sample period (10s) during which the accelerator was actively processing.
Shown as percent |
gcp.gke.container.accelerator.memory_used
(count) | Total accelerator memory allocated in bytes.
Shown as byte |
gcp.gke.container.accelerator.memory_total
(count) | Total accelerator memory in bytes.
Shown as byte |
gcp.gke.container.accelerator.memory_bandwidth_utilization
(count) | Current percentage of the accelerator memory bandwidth that is being used.
Shown as percent |
Eventos
La integración Google Cloud TPU no incluye eventos.
Checks de servicios
La integración Google Cloud TPU no incluye checks de servicio.
Solucionar problemas
¿Necesitas ayuda? Ponte en contacto con el servicio de asistencia de Datadog.
