Using Pipelining for TCP Connection

Pipelining is a technique used by NCache to optimize network communication by grouping multiple commands into a single batch and sending them over a TCP connection in one call. This technique reduces the overhead of sending multiple requests one by one and waiting for the acknowledgment. It also makes resource utilization more effective and speeds them up, especially for large data units. Pipelining provides a major performance boost for high performing machines.

Important

Pipelining is not configurable from NCache 5.3.5 onwards.

Pipelining for TCP Connection

NCache lets you send a chunk of multiple commands together in a defined time period, reducing the TCP overhead. The time interval of sending the chunk together is configurable by the user and after each interval, the commands are gathered and sent over the network. The server manages Pipelining settings for each connected client and propagates these configurations as needed. One cache can have one configuration for Pipelining which will connect to all the servers.

The minimum time for the batch interval is 50 µs and the maximum size for the batch interval is 5000 µs. The default value is 250 µs and it is enabled by default. Pipelining can only be used for clustered caches.

You can configure Pipelining using the NCache Management Center. Moreover, you can also configure Pipelining through the config file (config.ncconf).

See Also

Configure Pipelining
Data Load Balancing
Cache Server Backward Compatibility
Data Compression