- Teradata connector for Teradata Database
- Hive connector for CDH and HDP
- Presto connector for Presto, CDH, or HDP
- Spark SQL connector for CDH and HDP
- Oracle connector for Oracle Database
|System (Data Source)||
|Teradata Viewpoint||Viewpoint 16.10, 16.20 and later. The QueryGrid portlet package must be installed on the Viewpoint server.
The Viewpoint 16.10 edition must support multiple-system monitoring. QueryGrid 2.0x cannot be configured on Viewpoint editions that support only single-system monitoring.
|Kerberos||For environments secured with Kerberos, the most current version of Kerberos available.|
|Java||The latest version of Java 8 on the driver node or nodes in an initiating system (data source). A driver node invokes the connector method (for example, Teradata Database) used for a query on a target system (data source).
The Teradata/Oracle JDK is preferred.
|TMS, Server, or VM for QueryGrid Manager||One or more (separate server for each QueryGrid Manager instance).
Important: Do not install QueryGrid Manager on existing TMS servers hosting Viewpoint, Data Mover, Ecosystem Manager, and so forth, unless those products are uninstalled and the server is re-purposed to be a QueryGrid Manager server.
See QueryGrid Manager planning section below.
|Network||Network connectivity between the Teradata nodes and Hadoop master node, and all Hadoop data nodes, through customer LAN, BYNET, or InfiniBand.|
Number of QueryGrid Manager Instances
For each Data Center, determine the number of QueryGrid Managers needed. One instance of QueryGrid Manager is installed on one TMS or customer-supplied server, or VM.
Use the following guidelines. The guidelines are based on the minimum hardware requirements for the TMS, server, or VM.
|Per Data Center||Description|
|Minimum number of QueryGrid Manager instances||One instance for approximately every 200 Teradata QueryGrid-connected nodes
|Minimum number of QueryGrid Manager instances for high availability (HA)||Two instances
|Number of QueryGrid Manager instances to keep communication, heartbeats, query metrics, and log data local to a Data Center||At least one per Data Center, more if the number of nodes in the Data Center exceed the resource requirements of the single QueryGrid Manager instance|
Ensure that network ports are open between QueryGrid-attached systems (data sources) and QueryGrid Manager. Systems include the Teradata Database nodes, Hadoop master node, all Hadoop data nodes, and Presto nodes.
The following illustration shows the port numbers used between QueryGrid-attached systems and QueryGrid Manager in a fabric.
|1025||One-way connection from the target Teradata connector (driver) to the target Teradata Database systems (JDBC connection).|
|5000-5001||Two-way connections between all QueryGrid-attached nodes; other available ports can be used, but ports 5000-5001 are recommended.|
|7000-7001||Two-way connections between clustered Teradata QueryGrid Managers.|
|8080||One-way connection from the target Presto connector (driver) to the target Presto coordinator.|
|8443||One-way connection from the target Hive connector (driver) to the Apache Knox Gateway if Apache Knox Gateway is enabled on Hadoop nodes.|
|9300-9303||Two-way connections between clustered Teradata QueryGrid Managers.|
|9444||One-way HTTPS connection from QueryGrid-attached nodes to Teradata QueryGrid Manager.|
|9445||Two-way HTTPS connection between clustered Teradata QueryGrid Managers in a fabric.|
|10000||One-way connection between the target Hive connector (driver) to target Hiveserver 2.|
|10016||One-way connection from the target Spark SQL connector (driver) to the target Spark Thrift Server.|