腾讯云可观测平台任务式建模

一、命名空间

Namespace = QCE/TI_TRAINTASK

二、CPU

指标名	指标中文名	说明	单位	维度	统计规则
Cpuutil	CPU 利用率	CPU 利用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Gpuutil	GPU 利用率	GPU 利用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Instancecpuutil	CPU 利用率	CPU 利用率	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Instancegpuutil	GPU 利用率	GPU 利用率	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuUtil	GPU 使用率	GPU 使用率	%	appid insid subuin taskid uuid	[ 10s, last ] [ 60s, avg ] [ 300s, avg ]
JobCpuUtil	CPU 利用率	CPU 利用率	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuUtil	GPU 利用率	GPU 利用率	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]

三、GPU

指标名	指标中文名	说明	单位	维度	统计规则
Gpumemutil	GPU 显存利用率	GPU 显存利用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Gpumemvalue	显存使用量	显存使用量	MBytes	appid subuin taskid	[ 10s, sum ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Gpuutilmax	GPU 峰值利用率	GPU 峰值利用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, max ] [ 3600s, max ] [ 86400s, max ]
Instancegpumemutil	GPU 显存利用率	GPU 显存利用率	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Instancegpumemvalue	显存使用量	显存使用量	MBytes	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DcgmFiDevFbUsed	显存使用量	显存使用量	MBytes	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DcgmFiDevGpuUtil	GPU 使用率	GPU 使用率	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DcgmFiDevMemCopyUtil	显存带宽使用率	显存带宽使用率	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Fp16EngineActivity	FP16 活跃时间比	FP16 活跃时间比	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Fp32EngineActivity	FP32 活跃时间比	FP32 活跃时间比	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Fp64EngineActivity	FP64 活跃时间比	FP64 活跃时间比	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuDecUtil	GPU 解码使用率	GPU 解码使用率	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuEncUtil	GPU 编码器使用率	GPU 编码器使用率	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuFp16EngineActivity	FP16 活跃时间比	FP16 活跃时间比	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuFp32EngineActivity	FP32 活跃时间比	FP32 活跃时间比	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuFp64EngineActivity	FP64 活跃时间比	FP64 活跃时间比	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuMemoryClock	GPU 显存频率	GPU 显存频率	s	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuMemoryFree	GPU 显存空闲量	GPU 显存空闲量	MBytes/s	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuMemoryUtil	GPU 显存带宽使用率	GPU 显存带宽使用率	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuNvlinkBandwidth	nvlink 传输速率	nvlink 传输速率	GBytes/s	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuNvlinkRxMb	nvlink 接收数据量	nvlink 接收数据量	Mbps	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuNvlinkTxMb	nvlink 发送数据量	nvlink 发送数据量	Mbps	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuPcieBandwidth	PCIe 总线传输速率	PCIe 总线传输速率	GBytes/s	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuPcieRxMb	pcie 接收数据量	pcie 接收数据量	Mbps	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuPcieTxMb	pcie 发送数据量	pcie 发送数据量	Mbps	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuSmActivity	SM 活跃状态时间比	SM 活跃状态时间比	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuSmClock	SM 时钟频率	SM 时钟频率	s	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuTensorActivity	Tensor 活跃状态时间比	Tensor 活跃状态时间比	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobFp16EngineActivity	FP16 活跃时间比	FP16 活跃时间比	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobFp32EngineActivity	FP32 活跃时间比	FP32 活跃时间比	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobFp64EngineActivity	FP64 活跃时间比	FP64 活跃时间比	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuDecUtil	GPU 解码使用率	GPU 解码使用率	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuEncUtil	GPU 编码器使用率	GPU 编码器使用率	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuMemUtil	GPU 显存利用率	GPU 显存利用率	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuMemValue	显存使用量	显存使用量	MBytes	appid jobid subuin	[ 10s, sum ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuMemoryClock	GPU 显存频率	GPU 显存频率	s	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuMemoryFree	GPU 显存空闲量	GPU 显存空闲量	MBytes	appid jobid subuin	[ 10s, sum ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuMemoryUtil	显存使用率	显存使用率	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuNvlinkRxMb	nvlink 接收数据量	nvlink 接收数据量	Mbps	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuNvlinkTxMb	nvlink 发送数据量	nvlink 发送数据量	Mbps	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuPcieRxMb	pcie 接收数据量	pcie 接收数据量	Mbps	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuPcieTxMb	pcie 发送数据量	pcie 发送数据量	Mbps	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuSmClock	SM 时钟频率	SM 时钟频率	s	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobNvlinkBandwidth	nvlink 传输速率	nvlink 传输速率	Bytes/s	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobPcieBandwidth	PCIe 总线传输速率	PCIe 总线传输速率	Bytes/s	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobSmActivity	SM 活跃状态时间比	SM 活跃状态时间比	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobTensorActivity	Tensor 活跃状态时间比	Tensor 活跃状态时间比	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
NvlinkBandwidth	nvlink 传输速率	nvlink 传输速率	Bytes/s	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
PcieBandwidth	PCIe 总线传输速率	PCIe 总线传输速率	Bytes/s	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
SmActivity	SM 活跃状态时间比	SM 活跃状态时间比	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskFp16EngineActivity	FP16 活跃时间比	FP16 活跃时间比	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskFp32EngineActivity	FP32 活跃时间比	FP32 活跃时间比	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskFp64EngineActivity	FP64 活跃时间比	FP64 活跃时间比	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuDecUtil	GPU 解码使用率	GPU 解码使用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuDecUtilGpu	GPU 解码使用率	GPU 解码使用率	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuEncUtil	GPU 编码器使用率	GPU 编码器使用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuEncUtilGpu	GPU 编码器使用率	GPU 编码器使用率	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuMemoryClock	GPU 显存频率	GPU 显存频率	s	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuMemoryClockGpu	GPU 显存频率	GPU 显存频率	s	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuMemoryFree	GPU 显存空闲量	GPU 显存空闲量	MBytes	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuMemoryFreeGpu	GPU 显存空闲量	GPU 显存空闲量	MBytes	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuMemoryUtil	GPU 显存带宽使用率	GPU 显存带宽使用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuMemoryUtilGpu	GPU 显存带宽使用率	GPU 显存带宽使用率	%	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuNvlinkRxMb	nvlink 接收数据量	nvlink 接收数据量	Mbps	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuNvlinkRxMbGpu	nvlink 接收数据量	nvlink 接收数据量	Mbps	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuNvlinkTxMb	nvlink 发送数据量	nvlink 发送数据量	Mbps	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuNvlinkTxMbGpu	nvlink 发送数据量	nvlink 发送数据量	Mbps	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuPcieRxMb	pcie 接收数据量	pcie 接收数据量	Mbps	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuPcieRxMbGpu	pcie 接收数据量	pcie 接收数据量	Mbps	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuPcieTxMb	pcie 发送数据量	pcie 发送数据量	Mbps	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuPcieTxMbGpu	pcie 发送数据量	pcie 发送数据量	Mbps	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuSmClock	SM 时钟频率	SM 时钟频率	s	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuSmClockGpu	SM 时钟频率	SM 时钟频率	s	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskNvlinkBandwidth	nvlink 传输速率	nvlink 传输速率	Bytes/s	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskPcieBandwidth	PCIe 总线传输速率	PCIe 总线传输速率	Bytes/s	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskSmActivity	SM 活跃状态时间比	SM 活跃状态时间比	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskTensorActivity	Tensor 活跃状态时间比	Tensor 活跃状态时间比	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TensorActivity	Tensor 活跃状态时间比	Tensor 活跃状态时间比	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]

四、GPU

指标名	指标中文名	说明	单位	维度	统计规则
HcclCommAvgTime	HCCL 平均通信时间	HCCL 平均通信时间	us	appid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclCommCallNum	HCCL 调用次数	HCCL 调用次数	Count	appid insid subuin	[ 10s, last ] [ 60s, sum ] [ 300s, sum ] [ 3600s, sum ] [ 86400s, sum ]
HcclCommFinishedNum	HCCL 完成次数	HCCL 完成次数	Count	appid insid subuin	[ 10s, last ] [ 60s, sum ] [ 300s, sum ] [ 3600s, sum ] [ 86400s, sum ]
HcclCommMaxTime	HCCL 最大通信时间	HCCL 最大通信时间	us	appid insid subuin	[ 10s, max ] [ 60s, max ] [ 300s, max ] [ 3600s, max ] [ 86400s, max ]
HcclCommMinTime	HCCL 最小通信时间	HCCL 最小通信时间	us	appid insid subuin	[ 10s, min ] [ 60s, min ] [ 300s, min ] [ 3600s, min ] [ 86400s, min ]
HcclNetAvgBusbw	HCCL 平均总线带宽	HCCL 平均总线带宽	GBit/s	appid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclNetAvgLat	HCCL 平均时延	HCCL 平均时延	us	appid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobHcclCommAvgTime	HCCL 平均通信时间	HCCL 平均通信时间	us	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobHcclCommCallNum	HCCL 调用次数	HCCL 调用次数	Count	appid jobid subuin	[ 10s, sum ] [ 60s, sum ] [ 300s, sum ] [ 3600s, sum ] [ 86400s, sum ]
JobHcclCommFinishedNum	HCCL 完成次数	HCCL 完成次数	Count	appid jobid subuin	[ 10s, sum ] [ 60s, sum ] [ 300s, sum ] [ 3600s, sum ] [ 86400s, sum ]
JobHcclCommMaxTime	HCCL 最大通信时间	HCCL 最大通信时间	us	appid jobid subuin	[ 10s, max ] [ 60s, max ] [ 300s, max ] [ 3600s, max ] [ 86400s, max ]
JobHcclCommMinTime	HCCL 最小通信时间	HCCL 最小通信时间	us	appid jobid subuin	[ 10s, min ] [ 60s, min ] [ 300s, min ] [ 3600s, min ] [ 86400s, min ]
JobHcclNetAvgBusbw	HCCL 平均总线带宽	HCCL 平均总线带宽	GBit/s	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobHcclNetAvgLat	HCCL 平均时延	HCCL 平均时延	us	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]

五、RDMA

指标名	指标中文名	说明	单位	维度	统计规则
JobRdmaInpkt	RDMA 网卡入包量	RDMA 网卡入包量	pps	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobRdmaIntraffic	RDMA 网卡接收带宽	RDMA 网卡接收带宽	Mbps	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobRdmaOutpkt	RDMA 网卡出包量	RDMA 网卡出包量	pps	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobRdmaOuttraffic	RDMA 网卡发送带宽	RDMA 网卡发送带宽	Mbps	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
RdmaInpkt	RDMA 网卡入包量	RDMA 网卡入包量	pps	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
RdmaIntraffic	RDMA 网卡接收带宽	RDMA 网卡接收带宽	Mbps	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
RdmaOutpkt	RDMA 网卡出包量	RDMA 网卡出包量	pps	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
RdmaOuttraffic	RDMA 网卡发送带宽	RDMA 网卡发送带宽	Mbps	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskRdmaInpkt	RDMA 网卡入包量	RDMA 网卡入包量	pps	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskRdmaIntraffic	RDMA 网卡接收带宽	RDMA 网卡接收带宽	Mbps	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskRdmaOutpkt	RDMA 网卡出包量	RDMA 网卡出包量	pps	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskRdmaOuttraffic	RDMA 网卡发送带宽	RDMA 网卡发送带宽	Mbps	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]

六、cfs

指标名	指标中文名	说明	单位	维度	统计规则
CfsDataReadIoBytes	CFS 读带宽	CFS 读带宽	MBytes/s	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
CfsDataReadIoLatency	CFS 读延迟	CFS 读延迟	ms	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
CfsDataWriteIoBytes	CFS 写带宽	CFS 写带宽	MBytes/s	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
CfsDataWriteIoLatency	CFS 写延迟	CFS 写延迟	ms	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]

七、内存

指标名	指标中文名	说明	单位	维度	统计规则
Instancememutil	内存利用率	内存利用率	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Instancememvalue	内存使用量	内存使用量	MBytes	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Memutil	内存利用率	内存利用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
Memvalue	内存用量	内存用量	MBytes	appid subuin taskid	[ 10s, sum ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobMemUtil	内存利用率	内存利用率	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobMemValue	内存使用量	内存使用量	MBytes	appid jobid subuin	[ 10s, sum ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]

八、磁盘

指标名	指标中文名	说明	单位	维度	统计规则
DiskIoUtil	磁盘 ioutil	磁盘 ioutil	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DiskIoWait	磁盘 iowait	磁盘 iowait	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DiskReadByte	磁盘读取带宽	磁盘读取带宽	MBytes/s	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DiskReadIops	磁盘读取 iops	磁盘读取 iops	Count	appid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DiskUsageRadio	系统盘分区利用率	系统盘分区利用率	%	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DiskWriteByte	磁盘写入带宽	磁盘写入带宽	MBytes/s	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DiskWriteIops	磁盘写入 iops	磁盘写入 iops	Count	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobDiskIoUtil	磁盘 ioutil	磁盘 ioutil	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobDiskIoWait	磁盘 iowait	磁盘 iowait	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobDiskReadByte	磁盘读取带宽	磁盘读取带宽	MBytes/s	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobDiskReadIops	磁盘读取 iops	磁盘读取 iops	Count	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobDiskUsageRadio	系统盘分区利用率	系统盘分区利用率	%	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobDiskWriteByte	磁盘写入带宽	磁盘写入带宽	MBytes/s	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobDiskWriteIops	磁盘写入 iops	磁盘写入 iops	Count	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
NodeDiskLimit	节点磁盘总量	节点磁盘总量	GBytes	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
NodeDiskValue	节点磁盘使用量	节点磁盘使用量	GBytes	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
PodDiskLimit	实例磁盘总量	实例磁盘总量	GBytes	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
PodDiskValue	实例磁盘使用量	实例磁盘使用量	GBytes	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskDiskIoUtil	磁盘 ioutil	磁盘 ioutil	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskDiskIoWait	磁盘 iowait	磁盘 iowait	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskDiskReadByte	磁盘读取带宽	磁盘读取带宽	MBytes/s	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskDiskReadIops	磁盘读取 iops	磁盘读取 iops	Count	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskDiskUsageRadio	系统盘分区利用率	系统盘分区利用率	%	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskDiskWriteByte	磁盘写入带宽	磁盘写入带宽	MBytes/s	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskDiskWriteIops	磁盘写入 iops	磁盘写入 iops	Count	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]

九、其他

指标名	指标中文名	说明	单位	维度	统计规则
CfsStorageUsage	CFS 使用率	CFS 使用率	%	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
CfsStorageUsageGb	CFS 使用量	CFS 使用量	MBytes/s	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
CpuUsageCore	CPU 使用量	CPU 使用量	None	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
DcgmFiDevMemUtil	显存使用率	显存使用率	None	appid ins_gpu_num subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsClusterBytesReadRemote	goosefs 集群 Remote 读流量	goosefs 集群 Remote 读流量	MBytes/s	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsClusterCacheHit	goosefs 集群缓存命中率	goosefs 集群缓存命中率	%	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsClusterCapacityTotal	goosefs 集群缓存总容量	goosefs 集群缓存总容量	MBytes	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsClusterRemoteThroughput	goosefs 集群 Remote 读吞吐	goosefs 集群 Remote 读吞吐	Bytes/s	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsReadBytes	客户端读带宽	客户端读带宽	Bytes	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsReadLatency	客户端读延迟	客户端读延迟	ms	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsReadOps	客户端读 OPS	客户端读 OPS	Count/s	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsWriteBytes	客户端写带宽	客户端写带宽	Bytes	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsWriteLatency	客户端写延迟	客户端写延迟	ms	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsWriteOps	客户端写 OPS	客户端写 OPS	Count/s	appid fsid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsxNodeBytesRead	GooseFSx 读带宽	GooseFSx 读带宽	MBytes/s	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsxNodeBytesWritten	GooseFSx 写带宽	GooseFSx 写带宽	MBytes/s	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsxNodeReadLatency	GooseFSx 读延迟	GooseFSx 读延迟	ms	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GoosefsxNodeWriteLatency	GooseFSx 写延迟	GooseFSx 写延迟	ms	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuMemCopyUtil	GPU 显存带宽使用率	GPU 显存带宽使用率	None	appid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuPowerUsage	GPU 功耗	GPU 功耗	None	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
GpuTemperature	GPU 温度	GPU 温度	None	appid ins_gpu_num subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclAverageTime	HCCL 平均通信时间	HCCL 平均通信时间	us	AppId InstanceId SubUin Collective SizeScope	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclAvgBusbw	HCCL 平均总线带宽	HCCL 平均总线带宽	GBits	appid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclAvgLat	HCCL 平均时延	HCCL 平均时延	us	appid insid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclCallNum	HCCL 调用次数	HCCL 调用次数	Count/s	AppId InstanceId SubUin Collective SizeScope	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclFinishedNum	HCCL 完成次数	HCCL 完成次数	Count/s	AppId InstanceId SubUin Collective SizeScope	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclMaxTime	HCCL 最大通信时间	HCCL 最大通信时间	us	AppId InstanceId SubUin Collective SizeScope	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
HcclMinTime	HCCL 最小通信时间	HCCL 最小通信时间	us	AppId InstanceId SubUin Collective SizeScope	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobCpuUsageCore	CPU 使用量	CPU 使用量	None	appid jobid subuin	[ 10s, sum ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuMemCopyUtil	GPU 显存带宽使用率	GPU 显存带宽使用率	None	appid jobid subuin	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuPowerUsage	GPU 功耗	GPU 功耗	None	appid jobid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
JobGpuTemperature	GPU 温度	GPU 温度	None	appid jobid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
NetworkReceiveMbps	网络入流量	网络入流量	None	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
NetworkTransmitMbps	网络出流量	网络出流量	None	appid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskCpuUsageCore	CPU 使用量	CPU 使用量	None	appid subuin taskid	[ 10s, sum ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuMemCopyUtil	GPU 显存带宽使用率	GPU 显存带宽使用率	None	appid subuin taskid	[ 10s, avg ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TaskGpuNum	任务 GPU 卡数	任务 GPU 卡数	None	appid subuin taskid	[ 10s, last ] [ 60s, last ] [ 300s, last ] [ 3600s, last ] [ 86400s, last ]
TaskPendingDuration	任务排队时长	任务排队时长	None	appid subuin taskid	[ 10s, last ] [ 60s, last ] [ 300s, last ] [ 3600s, last ] [ 86400s, last ]
TurbocfsClientReadIoBytes	TurboCFS 节点读带宽	TurboCFS 节点读带宽	MBytes/s	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]
TurbocfsClientWriteIoBytes	TurboCFS 节点写带宽	TurboCFS 节点写带宽	MBytes/s	appid fsid insid subuin	[ 10s, last ] [ 60s, avg ] [ 300s, avg ] [ 3600s, avg ] [ 86400s, avg ]

十、各维度对应参数总览

参数名称	维度名称	维度解释	格式
Instances.N.Dimensions.0.Name	AppId	账号基本信息 APPID 的维度名称	输入 String 类型维度名称：AppId（SDK 调用时会自动获取，无需传参）
Instances.N.Dimensions.0.Value	AppId	账号基本信息 APPID	输入 ID，例如：1231231231（SDK 调用时会自动获取，无需传参）
Instances.N.Dimensions.1.Name	SubUin	子账号 ID 的维度名称	输入 String 类型维度名称：SubUin
Instances.N.Dimensions.1.Value	SubUin	子账号 ID	输入 ID，例如：100001231231
Instances.N.Dimensions.2.Name	InstanceId	训练任务实例 ID 的维度名称	输入 String 类型维度名称：InstanceId
Instances.N.Dimensions.2.Value	InstanceId	训练任务实例 ID（Pod 名）	输入具体实例 ID，例如：train-9187850047592xxxxx-9ludoo1s1n9c-master-0
Instances.N.Dimensions.3.Name	InstanceGpuNum	训练任务实例使用的 GPU 卡号（仅限 GPU 整卡任务）的维度名称	输入 String 类型维度名称：InstanceGpuNum
Instances.N.Dimensions.3.Value	InstanceGpuNum	训练任务实例使用的 GPU 卡号（仅限 GPU 整卡任务）	训练任务实例 ID 拼接 GPU 卡号/avg，输入具体实例 ID，例如：train-9187850047592xxxxx-9ludoo1s1n9c-master-0-0、train-9187850047592xxxxx-9ludoo1s1n9c-master-0-avg
Instances.N.Dimensions.4.Name	TaskId	训练任务的维度名称	输入 String 类型维度名称：TaskId
Instances.N.Dimensions.4.Value	TaskId	训练任务 ID	输入 ID，例如：train-9187850047592xxxxx
Instances.N.Dimensions.5.Name	JobId	训练任务实例（Job）的维度名称	输入 String 类型维度名称：JobId
Instances.N.Dimensions.5.Value	JobId	训练任务实例（Job）ID，对应指标表中的 Job 系列指标	输入 ID，例如：train-9187850047592xxxxx-9ludoo1s1n9c
Instances.N.Dimensions.6.Name	FsId	训练任务挂载文件系统的维度名称	输入 String 类型维度名称：FsId
Instances.N.Dimensions.6.Value	FsId	文件系统 ID（CFS / GooseFSx / TurboCFS），对应指标表中的 Cfs 系列、Goosefs 系列、Turbocfs 系列指标	输入文件系统 ID，例如：cfs-3a2b4c5d
Instances.N.Dimensions.7.Name	Uuid	实例唯一标识的维度名称	输入 String 类型维度名称：Uuid
Instances.N.Dimensions.7.Value	Uuid	实例唯一标识，对应卡维度 GPU 使用率 GpuUtil	输入实例唯一标识（实例重建后会变化），用于区分同名实例
Instances.N.Dimensions.8.Name	Collective	集合通信算子类型的维度名称	输入 String 类型维度名称：Collective
Instances.N.Dimensions.8.Value	Collective	集合通信算子类型	输入算子类型，例如：allreduce、allgather、broadcast、reducescatter
Instances.N.Dimensions.9.Name	SizeScope	集合通信消息大小分档的维度名称	输入 String 类型维度名称：SizeScope
Instances.N.Dimensions.9.Value	SizeScope	集合通信消息大小分档	输入分档，例如：0-1m、4g+

注意：说明：

注意：

十一、入参说明

·查询任务式建模指标监控数据，全部可用维度取值如下：

&Namespace=QCE/TI_TRAINTASK

&Instances.N.Dimensions.0.Name=AppId

&Instances.N.Dimensions.0.Value=具体的账号 ID

&Instances.N.Dimensions.1.Name=SubUin

&Instances.N.Dimensions.1.Value=具体的子账号 ID

&Instances.N.Dimensions.2.Name=InstanceId

&Instances.N.Dimensions.2.Value=训练任务实例 ID

&Instances.N.Dimensions.3.Name=InstanceGpuNum

&Instances.N.Dimensions.3.Value=训练任务实例使用的 GPU 卡号

&Instances.N.Dimensions.4.Name=TaskId

&Instances.N.Dimensions.4.Value=训练任务 ID

&Instances.N.Dimensions.5.Name=JobId

&Instances.N.Dimensions.5.Value=训练任务实例（Job）ID

&Instances.N.Dimensions.6.Name=FsId

&Instances.N.Dimensions.6.Value=文件系统 ID

&Instances.N.Dimensions.7.Name=Uuid

&Instances.N.Dimensions.7.Value=实例唯一标识

&Instances.N.Dimensions.8.Name=Collective

&Instances.N.Dimensions.8.Value=集合通信算子类型

&Instances.N.Dimensions.9.Name=SizeScope

&Instances.N.Dimensions.9.Value=集合通信消息大小分档

·按指标所属层级取用对应组合，例如：

·任务级：CPU 利用率

&Namespace=QCE/TI_TRAINTASK

&MetricName=Cpuutil

&Instances.N.Dimensions.0.Name=AppId

&Instances.N.Dimensions.0.Value=1231231231

&Instances.N.Dimensions.1.Name=SubUin

&Instances.N.Dimensions.1.Value=100001231231

&Instances.N.Dimensions.2.Name=TaskId

&Instances.N.Dimensions.2.Value=train-9187850047592xxxxx

·卡级：GPU 使用率

&Namespace=QCE/TI_TRAINTASK

&MetricName=DcgmFiDevGpuUtil

&Instances.N.Dimensions.0.Name=AppId

&Instances.N.Dimensions.0.Value=1231231231

&Instances.N.Dimensions.1.Name=SubUin

&Instances.N.Dimensions.1.Value=100001231231

&Instances.N.Dimensions.2.Name=InstanceGpuNum

&Instances.N.Dimensions.2.Value=train-9187850047592xxxxx-9ludoo1s1n9c-master-0-0

腾讯云可观测平台 任务式建模

一、命名空间

二、CPU

三、GPU

四、GPU

五、RDMA

六、cfs

七、内存

八、磁盘

九、其他

十、各维度对应参数总览

十一、入参说明

相关内容：

腾讯云可观测平台任务式建模