FAQ

• How are the different HDInsight cluster types billed?

  HDInsight deploys different number of nodes for each cluster type. Within a given cluster type, there are different roles for the various nodes, which allow a customer to size those nodes in a given role appropriate to the details of their workload. For example, a Hadoop cluster can have its worker nodes provisioned with a large amount of memory if the type of analytics being performed are memory intensive.

  HDInsights’ Hadoop Cluster can deploy three kinds of roles:

  • Head node (2 nodes)
  • Data node (at least 1 node)
  • Zookeeper nodes (3 nodes)

  HDInsight’s HBase Cluster can deploy three kinds of roles:

  • Control Server (2 nodes)
  • Zone Server (at least 1 node)
  • Main node/Zookeeper node (3 nodes)

  HDInsight’s Storm Cluster can deploy three kinds of roles:

  • Nimbus node (2 nodes)
  • Supervision Server (at least 1 node)
  • Zookeeper node (3 nodes)
• If my cluster ran for less than an hour, how much would I get billed?

  We charge for the number of minutes your cluster is running, rounded to the nearest minute, not hour.

• Could you give me an example on how billing works?

  If you run a cluster for 100 hours in US East with two D13 v2 head nodes, three D12 v2 data nodes, and three D11 v2 zookeepers, the billing would be the following in the scenario:

  On a Standard HDInsight cluster—100 hours x ( 2 x ￥7.8867/hour + 3 x ￥3.9434/hour + 3 x ￥1.9717/hour) = ￥3351.87

• How can I check that I have properly stopped an HDInsight cluster and that I am not being billed for it?

  In order to stop an HDInsight cluster, you must delete the cluster. By default, all data an HDInsight cluster operates on resides in Azure Blob storage, so data will not be impacted by this. If you want to preserve your Hive metadata (tables, schemas) you should provision a cluster with an external metadata store. You can find more details in this documentation .

• How many data nodes do I need for my HDInsight cluster?

  The number of data nodes will vary depending on your needs. With the elasticity available in Azure cloud services, you can try a variety of cluster sizes to determine your own optimal mix of performance and cost, and only pay for what you use at any given time. Clusters can also be scaled on demand to grow and shrink to match the requirements of your workload.

• What if I need more HDInsight data nodes than my subscription allows?

  Each subscription has a default limit on how many HDInsight data nodes can be created. If you need to create a larger HDInsight cluster or multiple HDInsight clusters that together exceed your current subscription maximum, you can request that your subscription's billing limits be increased. Please open "Support Type" for related operations. Depending on the maximum nodes per subscription that you request, you may be asked for additional information that will allow us to optimize your deployment(s).

• How much would a cluster with "x" data nodes cost?

  To estimate the cost of clusters of various sizes, try the Azure Calculator .

• How can I reduce costs on clusters I use infrequently?

  There are a number of options to reduce the costs:

  • Drive higher utilization of your existing clusters.

    1.Delete clusters while not in use. For more information about deleting a cluster, see Delete an HDInsight cluster using your browser, PowerShell, or the Azure CLI

    2.Scale down. For more information about manually scaling clusters, see Scale HDInsight clusters

  • Deploy the clusters with lower cost. This includes proper planning on how many nodes to use, which type of node to use for head nodes and worker nodes, and which region to launch the cluster as HDInsight offers many different node types to deploy to, with a range of pricing options. Review the Base price/node-hour section of this article for pricing and for more information see Capacity planning for HDInsight clusters
• How much would a cluster with "x" data nodes cost?

  To estimate the cost of clusters of various sizes, try the Azure Calculator .