Requirements to Deploy a Windows HPC Cluster with Remote Databases
This topic lists the requirements for a computer where a remote HPC database can be installed (the remote database server). It also lists the requirements for the connection between the head node computer and each remote database server.
In this topic
Requirements for the remote database server
This section lists the requirements for each computer that you want to use as a remote database server. It is especially important to review these requirements if your organization has a database infrastructure that is centrally managed, where specific security policies might be implemented. Some of the requirements for installing HPC databases remotely might not comply with those security policies.
Each computer where you want to install an HPC database must meet the following requirements:
A version of SQL Server that is supported by your version of HPC Pack must be installed and running on the computer. For information about supported versions of SQL Server, see System Requirements for HPC Pack 2016.
Important
- The Express edition of SQL Server is supported; however, it restricts the sizes of the databases and is not recommended for HPC clusters that have large numbers of nodes. For example, if you decide to install the Express edition of SQL Server 2016 and you are deploying a large cluster, be aware that its database size limitation is 10 GB per database.
- The Compact edition of SQL Server is not supported.
Note
If you are unsure about what version or what edition of SQL Server you have installed, see How to identify your SQL Server version and edition.
The account used by each SQL Server service must be properly configured, as follows:
If you are deploying a single HPC Pack head node, it is recommended that you assign a built-in account (Local System, Local Service, or Network Service) to each SQL Server service.
If you are deploying a domain-joined HPC Pack head node in the context of a service fabric cluster or failover cluster, you must assign a domain user account to each SQL Server service.
If you are deploying a non-domain-joined HPC Pack headnode, you must use SQL Server authentication to each SQL Server service.
One or more instances of SQL Server must be available on the computer, and these instances can be used to create the HPC databases.
A login is created on each instance for the domain account or accounts that are designated to install, reinstall, update, or upgrade HPC Pack.
The appropriate role is assigned to all logins that are created for the HPC cluster, as follows:
- For HPC Pack 2012 and later: the db_datareader, db_datawriter, and db_ddladmin database-level roles.
Each instance is configured to be accessed remotely.
Firewall exceptions are created on the computer to enable remote access to the databases.
The computer has storage and processing resources to efficiently handle the new databases.
Requirements for the network connection between the head node computer and the remote database server
This section lists the requirements of the network connection between the computer where you install HPC Pack head node and each remote database server.
The network connection must meet the following requirements:
The head node computer and the remote database servers must be connected to the same domain or to trusted domains if they are domain joined.
The head node computer must be connected to all the HPC networks.
Requirements for the network connection between the head node computer and Azure SQL databases
This section lists the requirements of the network connection between the computer where you install HPC Pack head node and Azure SQL databases.
The network connection must meet the following requirements:
The head node computer must be able to connect to the Azure SQL databases. You shall configure the firewall rules for Azure SQL databases correctly to make sure the connection from head node is not blocked.
The head node computer must be connected to all the HPC networks.