RFA C++ buffer settings

SMR

Hello Team,

We have RFA C++ ver. 7.x application subscribing to 10K RICs. Several times during the market hours the application logs misses updates due to buffer overflow (logged internally by the application).

Are the below [numOutputBuffers] setting enough for high volume data updates? What other parameters can be tuned to optimize the application processing of updates?

The RFA configuration is as below

\Sessions\Session0\connectionList = "Connection_0"

\Sessions\Session0\waitTimeout = 3600000

\Sessions\Session0\recoveryTimeout = 18000000

\Sessions\Session0\recoveryQueueUpperLimit = 40000

\Sessions\Session0\responseQueueMaxBatchSize = 500

\Connections\Connection_0\connectionType = "SSLED"

\Connections\Connection_0\serverList = "fp_ads1, fp_ads2"

\Connections\Connection_0\portNumber = 8101

\Connections\Connection_0\userName = "feedpricer_956"

\Connections\Connection_0\marketDataItemSubTimeout = 12375000

\Connections\Connection_0\connectRetryInterval = 5000

\Connections\Connection_0\numOutputBuffers = 100

\Connections\Connection_0\messageThrottleRate = 150

Thank you.

Regards

Jirapongse

@SMR

From the description, the server cut the connection due to the buffer overflow condition. Therefore, the number of output buffers must be increased on the server, not the client.

The buffer overflow condition could be from a network issue or a slow consumer.

If the application is a slow consumer, you need to modify the application to process retrieved messages faster. Otherwise, you can create multiple event queues and use multiple threads to dispatch events in the event queues, or you can use RFA’s Horizontal Scaling feature to create multiple instances of consumers on multi-core processors.

For more information regarding Horizontal Scaling, please refer to the RFA Developers Guide.