NodeType (NODE) connection parameter

Separately licensed component required

Read-only scale-out and database mirroring each require a separate license. See Separately licensed components.

Controls load balancing among database servers involved in read-only scale-out.

Syntax

{ NodeType | NODE }={ DIRECT | PRIMARY | COPY }

Allowed values

DIRECT This is the default setting. When the NodeType connection parameter is set to DIRECT, the database server accepts the connection without performing load balancing or redirection.
PRIMARY If the NodeType connection parameter is set to PRIMARY and you have connected to the primary server, the connection is accepted. If you have connected to the mirror server or a copy node, the database server redirects the connection to the primary server.
COPY When the NodeType connection parameter is set to COPY, the database server examines the copy nodes in its own branch (including itself if it is not the root node) and chooses the copy node with the lightest load. If the database server does not choose itself, it redirects the client to the chosen database server.

This connection parameter is used by client applications to control load balancing among database servers involved in read-only scale-out. Clients specify this connection parameter to indicate what type of database server they want to connect to. The database server that the client initially connects to decides which database server should handle the connection, and if necessary, redirects the connection to the appropriate database server by returning the address of the database server. The client is disconnected from the original database server, and then connects to the appropriate database server.

When you specify COPY, the load balancing occurs on the branch that contains the database server the client connected to. A branch consists of that database server and any of its children. This feature may be useful for performing load balancing among geographically separate servers.

If a client connection is redirected to a different database server, the value specified in the ServerName (Server) connection parameter and the value of the Name database server property do not match. See Name server property.

When NodeType is specified, the application typically connects to the root node, and that database server determines which copy node is least heavily loaded. The connection is then redirected to that copy node. If the application makes and drops several such connections within a short period of time, the connection is pooled and the root node is not asked which copy server to use. Using connection pooling reduces the load on the root node, but may not give expected behavior. The application can specify that its connections cannot be pooled to ensure that the root server determines which copy node to connect on each connection.

Example

The primary and mirror servers are located in Waterloo, Ontario. Clients running in Japan may want to set up a group of copy nodes in Japan to reduce connection latency. If a branch of copy nodes in Japan was a direct child of the root node, then other children could be added beneath the copy node in Japan. Clients would connect to the hub and specify NodeType=COPY. Japanese local, read-only connections would then be load-balanced among the local database servers, and no client would ever connect to database servers elsewhere. If a client needed to make changes to the database, they would change their connection string to use NodeType=PRIMARY.

Discuss this page in DocCommentXchange.