Set the Number of Azure Proxy Nodes
Applies To: Microsoft HPC Pack 2008 R2, Microsoft HPC Pack 2012, Microsoft HPC Pack 2012 R2
Starting in HPC Pack 2012, you can configure the number of proxy node instances that are created automatically to support an Azure node deployment. These proxy nodes are Azure worker role instances that facilitate and load-balance the communication between on-premises head nodes and the cluster nodes that are hosted in Azure. The proxy nodes are created automatically at the same time that Azure nodes are provisioned, and they are removed automatically when the Azure nodes are stopped. The default number of proxy nodes is 2.
You can specify that the deployment uses a fixed number of proxy nodes, or that the deployment uses a number of proxy nodes that varies with the size of the deployment. For example, you can specify that an Azure node deployment uses 2 proxy nodes for up to 400 Azure nodes, and adds 1 proxy node per additional 200 Azure nodes.
Note
In HPC Pack 2008 R2 (with Service Pack 1 or later), 2 proxy nodes (size Small) are created when you provision the Azure nodes. The number of proxy nodes is not configurable by using the management tools in HPC Pack 2008 R2. However, the number of proxy nodes in an existing Azure node deployment can be increased or decreased by using the tools in the Azure Management Portal.
General recommendations
To avoid a single point of failure, configure at least 2 proxy nodes.
You should configure a number of proxy nodes that will support the number of Azure nodes that you plan to add to the cluster and the workloads that will be running on the cluster. For typical clusters and workloads, a minimum of 1 proxy node per 200 Azure nodes is recommended. Certain job types, such as MPI jobs and parametric sweep jobs, can place greater demands on the proxy nodes. If the number of proxy nodes is too small, the proxy nodes can be a bottleneck for communication between the on-premises cluster nodes and the Azure nodes. If the number of proxy nodes is too large, you may be using unnecessary Azure resources.
To help determine the amount of traffic on the proxy nodes, you can monitor the proxy nodes by using the cloud service monitoring tools in the Azure Management Portal. By default you can monitor metrics on the proxy role instances such as CPU utilization and network traffic. For example, if the CPU utilization or network traffic is found to be high, you can add proxy nodes. To do this, you can increase the number of proxy instances in the Azure node template, and then redeploy the Azure nodes or add nodes to an existing deployment. Alternatively, you can scale up the number of proxy instances in an existing deployment by using the tools in the Azure Management Portal. For more information, see How to Scale a Cloud Service and Linked Resources.
Additional considerations
The proxy role instances are not listed in HPC Cluster Manager after the Azure nodes are provisioned. However, the instances appear in the Azure Management Portal, and the number of proxy instances can also be managed there. The instances are named HPCProxy_IN_0, HPCProxy_IN_1, and so on.
The proxy role instances incur charges in Azure along with the Azure node instances.
The proxy role instances consume cores that are allocated to the subscription, and correspondingly reduce the number of cores that are available to deploy Azure nodes.
Starting in HPC Pack 2012, the size of the proxy role instances is set according to the size of the Azure nodes in a deployment. See the following table.
Note
Availability of Azure node sizes varies, depending on the version of HPC Pack.
Node size
Proxy role instance size
Small, Medium, Large, Extra Large, A5, A6, A7
Medium
A8, A9
A8