Stay organized with collections
Save and categorize content based on your preferences.
This page explains the basic terminology and concepts of pipeline processing in
Cloud Data Fusion.
Pipeline performance depends on the following aspects:
The size and characteristics of your data
The structure of your pipeline
Cluster sizing
Plugins that your Cloud Data Fusion pipeline uses
Pipeline processing terminology in Cloud Data Fusion
The following terminology applies in pipeline processing in
Cloud Data Fusion.
Machine type
Type of virtual machines (VMs) used (CPU, memory).
Cluster
A group of VMs working together to handle large-scale data processing tasks.
Master and worker nodes
Physical or virtual machines that can do processing. Master nodes usually
coordinate work. Worker nodes run executors that process data. They have
machine characteristics (amount of memory and number of vCores available for
processes).
vCores, Cores, or CPUs
A resource that does computing. Usually your nodes provide a certain amount of
Cores and your Executors request one or a few CPUs. Balance this along with
memory, or you might underutilize your cluster.
Driver
A single VM that acts as the central coordinator for the entire cluster. It
manages tasks, schedules work across worker nodes, and monitors job progress.
Executors
Multiple VMs performing the actual data processing tasks, as instructed by the
driver. Your data is partitioned and distributed across these executors for
parallel processing. To utilize all of the executors, you must have enough
splits.
Splits or partitions
A dataset is split into splits (other name partitions) to process data in
parallel. If you don't have enough splits, you can't utilize the whole
cluster.
Performance tuning overview
Pipelines are executed on clusters of machines. When you choose to run
Cloud Data Fusion pipelines on Dataproc clusters (which is the
recommended provisioner), it uses YARN (Yet Another Resource Negotiator) behind
the scenes. Dataproc utilizes YARN for resource management within
the cluster. When you submit a Cloud Data Fusion pipeline to a
Dataproc cluster, the underlying Apache Spark job leverages YARN
for resource allocation and task scheduling.
A cluster consists of master and worker nodes. Master nodes are generally
responsible for coordinating work, while worker nodes perform the actual work.
Clusters will normally have a small number of master nodes (one or three) and a
large number of workers. YARN is used as the work coordination system. YARN runs
a Resource Manager service on the master node and a Node Manager service on each
worker node. Resource Managers coordinate amongst all the Node Managers to
determine where to create and execute containers on the cluster.
On each worker node, the Node Manager reserves a portion of the available
machine memory and CPUs for running YARN containers. For example, on a
Dataproc cluster, if your worker nodes are n1-standard-4 VMs
(4 CPU, 15 GB memory), each Node Manager will reserve 4 CPUs and
12 GB memory for running YARN containers. The remaining 3 GB of memory
is left for the other Hadoop services running on the node.
When a pipeline is run on YARN, it will launch a pipeline workflow driver, a
Spark driver, and many Spark executors in Dataproc.
The workflow driver is responsible for launching the one or more Spark programs
that make up a pipeline. The workflow driver usually doesn't do much work. Each
Spark program runs a single Spark driver and multiple Spark executors. The
driver coordinates work amongst the executors, but usually doesn't perform any
actual work. Most of the actual work is performed by the Spark executors.
[[["Easy to understand","easyToUnderstand","thumb-up"],["Solved my problem","solvedMyProblem","thumb-up"],["Other","otherUp","thumb-up"]],[["Hard to understand","hardToUnderstand","thumb-down"],["Incorrect information or sample code","incorrectInformationOrSampleCode","thumb-down"],["Missing the information/samples I need","missingTheInformationSamplesINeed","thumb-down"],["Other","otherDown","thumb-down"]],["Last updated 2025-08-29 UTC."],[[["\u003cp\u003eThis page introduces pipeline processing concepts and terminology within Cloud Data Fusion, emphasizing how data pipelines are executed.\u003c/p\u003e\n"],["\u003cp\u003ePipeline performance is influenced by data size, pipeline structure, cluster sizing, and the plugins used.\u003c/p\u003e\n"],["\u003cp\u003eKey components in pipeline processing include machine types, clusters, master/worker nodes, vCores, drivers, executors, and splits, each playing a specific role in data processing.\u003c/p\u003e\n"],["\u003cp\u003eCloud Data Fusion pipelines on Dataproc clusters use YARN for resource management, coordinating work across master and worker nodes.\u003c/p\u003e\n"],["\u003cp\u003eThe actual data processing work is performed by Spark executors, which are coordinated by a Spark driver, with the workflow driver managing the overall pipeline execution.\u003c/p\u003e\n"]]],[],null,["# Pipeline performance overview\n\nThis page explains the basic terminology and concepts of pipeline processing in\nCloud Data Fusion.\n\nPipeline performance depends on the following aspects:\n\n- The size and characteristics of your data\n- The structure of your pipeline\n- Cluster sizing\n- Plugins that your Cloud Data Fusion pipeline uses\n\nPipeline processing terminology in Cloud Data Fusion\n----------------------------------------------------\n\nThe following terminology applies in pipeline processing in\nCloud Data Fusion.\n\nMachine type\n: Type of virtual machines (VMs) used (CPU, memory).\n\nCluster\n: A group of VMs working together to handle large-scale data processing tasks.\n\nMaster and worker nodes\n: Physical or virtual machines that can do processing. Master nodes usually\n coordinate work. Worker nodes run executors that process data. They have\n machine characteristics (amount of memory and number of vCores available for\n processes).\n\nvCores, Cores, or CPUs\n: A resource that does computing. Usually your nodes provide a certain amount of\n Cores and your Executors request one or a few CPUs. Balance this along with\n memory, or you might underutilize your cluster.\n\nDriver\n: A single VM that acts as the central coordinator for the entire cluster. It\n manages tasks, schedules work across worker nodes, and monitors job progress.\n\nExecutors\n: Multiple VMs performing the actual data processing tasks, as instructed by the\n driver. Your data is partitioned and distributed across these executors for\n parallel processing. To utilize all of the executors, you must have enough\n splits.\n\nSplits or partitions\n: A dataset is *split* into splits (other name partitions) to process data in\n parallel. If you don't have enough splits, you can't utilize the whole\n cluster.\n\nPerformance tuning overview\n---------------------------\n\nPipelines are executed on clusters of machines. When you choose to run\nCloud Data Fusion pipelines on Dataproc clusters (which is the\nrecommended provisioner), it uses YARN (Yet Another Resource Negotiator) behind\nthe scenes. Dataproc utilizes YARN for resource management within\nthe cluster. When you submit a Cloud Data Fusion pipeline to a\nDataproc cluster, the underlying Apache Spark job leverages YARN\nfor resource allocation and task scheduling.\n\nA cluster consists of master and worker nodes. Master nodes are generally\nresponsible for coordinating work, while worker nodes perform the actual work.\nClusters will normally have a small number of master nodes (one or three) and a\nlarge number of workers. YARN is used as the work coordination system. YARN runs\na Resource Manager service on the master node and a Node Manager service on each\nworker node. Resource Managers coordinate amongst all the Node Managers to\ndetermine where to create and execute containers on the cluster.\n\nOn each worker node, the Node Manager reserves a portion of the available\nmachine memory and CPUs for running YARN containers. For example, on a\nDataproc cluster, if your worker nodes are n1-standard-4 VMs\n(4 CPU, 15 GB memory), each Node Manager will reserve 4 CPUs and\n12 GB memory for running YARN containers. The remaining 3 GB of memory\nis left for the other Hadoop services running on the node.\n\nWhen a pipeline is run on YARN, it will launch a pipeline workflow driver, a\nSpark driver, and many Spark executors in Dataproc.\n\nThe workflow driver is responsible for launching the one or more Spark programs\nthat make up a pipeline. The workflow driver usually doesn't do much work. Each\nSpark program runs a single Spark driver and multiple Spark executors. The\ndriver coordinates work amongst the executors, but usually doesn't perform any\nactual work. Most of the actual work is performed by the Spark executors.\n\nWhat's next\n-----------\n\n- Learn about [parallel processing](/data-fusion/docs/concepts/parallel-processing) in Cloud Data Fusion."]]
