E-MapReduce:Query profile

Execution overview

The Execution Details page shows a visual execution plan tree on the left. Each operator node is labeled with its execution time and percentage. Nodes that consume a lot of time, such as HASH_JOIN, are highlighted in red. The lines between nodes indicate the number of rows processed. On the right, the top section shows a sorted list of nodes by time percentage, and the bottom section provides an overview of summary metrics.

The Execution Details page displays key summary metrics, including execution time, I/O performance, and network data transfer volume.

• Total

  Metric	Description
  Total	The total time consumed by the query. This includes the duration of the planning, execution, and analysis phases.
  PlannerTotalTime	The planning time. This includes parsing the SQL, analysis, optimization, and generating the execution plan.
  ExecutionWallTime	The query execution time and the BE-side execution time (wall-clock time of execution).
  CollectProfileTime	The profile collection time. This is the time spent collecting profile metrics from BEs.

• Execution time

  Metric	Description
  ExecutionWallTime	The total time consumed by the query execution.
  CumulativeCpuTime	The sum of CPU time for all BE nodes. This reflects the total CPU resource consumption.
  CpuUtilization	CPU utilization = CumulativeCpuTime / ExecutionWallTime. This indicates the average number of CPU cores used.
  CumulativeWaitTime	The cumulative wait time of the pipeline. This includes scheduling wait time (ScheduleTime) and blocking time (PendingTime) due to conditions such as an empty input queue, a full output queue, or unready dependencies.
  OperatorCumulativetime	The sum of the execution times for all operators. This includes I/O, network, and computation.
  IO	The cumulative I/O time for all Scan operators. This includes local disk reads, remote reads, and data cache access.

• I/O

  Metric	Description
  RawRowsRead	The total number of records scanned by all SCAN nodes.
  DiskReadBytes	The total amount of compressed data read by all SCAN execution nodes.
  LocalDiskReadBytes	The total size of compressed data read from the local cache by all Connector Scan execution nodes. This metric applies only to shared-data instances.
  RemoteReadBytes	The total size of compressed data read from OSS by all Connector Scan execution nodes. This metric applies only to shared-data instances.
  ResultRows	The total number of records output by all SCAN execution nodes.
  ResultBytes	The total amount of data read by all SCAN execution nodes.

• Network

  Metric	Description
  Bytes sent over network	The total number of bytes transferred by all Exchange execution nodes. This corresponds to the BytesSent metric.

Operator

Query level

• Summary metrics

  Metric	Description
  Total	The query's total elapsed time, including time spent in the Planning, Executing, and Profiling phases.
  QueryCpuCost	The cumulative CPU time consumed by the query. This value is the sum of CPU time from all concurrent processes, so it can exceed the actual execution time.
  QueryMemCost	The total memory consumed by the query.
  Variables	Variables associated with the query.

• Pipeline level

  Metric	Description
  ActiveTime	The driver execution time.
  DriverTotalTime	The total time consumed by the driver.
  PendingTime	The time a driver waits for its inputs and prerequisites.

Operator level

• General operator metrics

  Metric	Description
  OperatorTotalTime	Total execution time of the operator.
  PushRowNum	Total rows output by the operator.
  PullRowNum	Total rows input to the operator.
  PullChunkNum	Total chunks input to the operator.
  PushChunkNum	Total chunks output by the operator.
  PeakMemoryUsage	The operator's peak memory usage.

• OLAP Scan Operator

  Metric	Description
  Table	The table name.
  ScanTime	Total time spent on scan operations, which are performed in an asynchronous I/O thread pool.
  TabletCount	Number of tablets scanned.
  PushdownPredicates	The number of pushed-down predicates.
  BytesRead	The amount of data read.
  CompressedBytesRead	The amount of compressed data read.
  IOTime	Total I/O time.
  BitmapIndexFilterRows	The number of rows filtered by bitmap indexes.
  BloomFilterFilterRows	The number of rows filtered by the Bloom filter.
  SegmentRuntimeZoneMapFilterRows	The number of rows filtered by the runtime zone map.
  SegmentZoneMapFilterRows	The number of rows filtered by the zone map.
  ShortKeyFilterRows	The number of rows filtered by the short key.
  ZoneMapIndexFilterRows	The number of rows filtered by the zone map index.

• Connector Scan Operator

  These metrics are in addition to those of the OLAP Scan Operator and apply to instances that use compute and storage separation.

  Metric	Description
  CompressedBytesReadLocalDisk	The amount of compressed data read from the local cache on the compute node.
  CompressedBytesReadRemote	The total amount of compressed data read from OSS.
  IOTimeLocalDisk	The time spent on I/O operations when reading data from the local cache.
  IOTimeRemote	The time spent on I/O operations when reading data from OSS.

• Exchange Operator

  • Sink

    Metric	Description
    PartType	The data distribution mode. Valid values: UNPARTITIONED, RANDOM, HASH_PARTITIONED, and BUCKET_SHUFFLE_HASH_PARTITIONED.
    BytesSent	The amount of data sent.
    OverallThroughput	The overall data transfer throughput.
    NetworkTime	The time spent transmitting packets, excluding post-reception processing time.
    WaitTime	The time spent waiting because the sender-side queue is full.
    NetworkBandwidth	The calculated network bandwidth during data transfer.

  • Source

    Metric	Description
    SenderWaitLockTime	Total time the sender waited for a lock.
    BytesReceived	The amount of data received.
    DecompressChunkTime	Total time spent decompressing data chunks.
    DeserializeChunkTime	Total time spent deserializing data chunks.
    SenderTotalTime	The total time spent by the sender.

• Aggregate Operator

  Metric	Description
  GroupingKeys	Columns used in the GROUP BY clause.
  AggregateFunctions	The aggregate functions applied.
  AggComputeTime	The time spent computing aggregate functions.
  ExprComputeTime	The time spent computing expressions.
  HashTableSize	Size of the hash table used for aggregation.

• Join Operator

  • Probe

    Metric	Description
    DistributionMode	The data distribution mode.
    JoinType	The join type.
    OtherJoinConjunctEvaluateTime	The time spent evaluating other join conjuncts.
    ProbeConjunctEvaluateTime	The time spent evaluating probe conjuncts.
    SearchHashTableTime	The time spent searching the hash table.
    WhereConjunctEvaluateTime	The time spent evaluating WHERE conjuncts.

  • Build

    Metric	Description
    JoinPredicates	Predicates used in the join condition.
    JoinType	The join type.
    BuildBuckets	Number of buckets in the join's hash table.
    BuildHashTableTime	Time spent building the hash table for the join.
    RuntimeFilterBuildTime	The time spent building the runtime filter.
    RuntimeFilterNum	The number of runtime filters.
    DistributionMode	The data distribution mode.

• Window Function Operator

  Metric	Description
  ComputeTime	The time spent computing window functions.
  PartitionKeys	Keys used to partition the window.
  AggregateFunctions	The aggregate functions applied.

• Sort Operator

  Metric	Description
  SortKeys	The sort keys.
  SortType	The sorting method used, such as full sort or top-N sort.
  MergingTime	Time spent merging sorted runs of data.
  SortingTime	The time spent sorting data.

• TableFunction Operator

  Metric	Description
  TableFunctionExecTime	The time spent computing the table function.
  TableFunctionExecCount	The number of times the table function was executed.

• Project Operator

  Metric	Description
  ExprComputeTime	The time spent computing expressions.
  CommonSubExprComputeTime	The time spent computing common subexpressions.

• LocalExchange Operator

  Metric	Description
  Type	The type of local exchange. Valid values: Passthrough, Partition, and Broadcast.
  ShuffleNum	The number of shuffles. This metric applies only when Type is Partition.