This topic describes the metrics for the Elastic Algorithm Service (EAS) provided by Machine Learning Platform for AI (PAI) for online inference.

  • Set the Namespace parameter to acs_learn.
  • Set the Period parameter to an integral multiple of 60s. The default value is 60s.
Metric Unit Metric Dimensions Statistics
CPUConsumption count cpu_core_usage userId and serviceName Average and Maximum
GPUUtilization % gpu_util userId and serviceName Average and Maximum
MemoryComsumption byte memory_rss userId and serviceName Average, Maximum, and Minimum
QueryPerSecondTotal frequency qps_total userId and serviceName Value, Average, and Maximum
ResponsePerSecondWithStatusCode2xx frequency rps_status_2xx userId and serviceName Average, Maximum, Value, and Minimum
2xxResponseRatio % rps_status_2xx_ratio userId and serviceName Average, Minimum, Maximum, and Value
ResponsePerSecondWithStatusCode4xx frequency rps_status_4xx userId and instanceId Average, Maximum, and Value
4xxResponseRatio % rps_status_4xx_ratio userId and serviceName Average, Maximum, Minimum, and Value
ResponsePerSecondWithStatusCode5xx frequency rps_status_5xx userId and serviceName Average, Maximum, and Value
5xxResponseRatio % rps_status_5xx_ratio userId and serviceName Average, Maximum, Minimum, and Value
ResponseTime microseconds rt userId and serviceName Average, Maximum, and Minimum
IngressTraffic bps traffic_in userId and serviceName Average, Maximum, Minimum, and Value
EgressTraffic bps traffic_out userId and serviceName Average, Maximum, Minimum, and Value