+

Search Tips   |   Advanced Search

The <maxConcurrentConnectionsPerNode> element of the HTTP adapter

The <maxConcurrentConnectionsPerNode> element defines the maximum number of concurrent HTTP connections from IMFP server to the back-end service.

MPF does not limit the incoming service requests from MobileFirst applications (this can be configured at the application server level) but only limits the number of concurrent HTTP connections to the back-end service.

The default number of concurrent HTTP connections is 50. It is recommended to modify this number based on the expected concurrent requests to the adapter and the maximum requests allowed on the back-end service. It is also recommended to configure the back-end service to limit the number of concurrent incoming requests.

Consider a two-node system, where the expected load on the system is 100 concurrent requests and the back-end service can support up to 80 concurrent requests. It is recommended to set maxConcurrentConnectionsPerNode to 40. This will make sure that no more than 80 concurrent requests are made to the back-end service.

If we increase the value, the back-end application needs more memory. To avoid memory issues, be careful not to set this value too high. Instead, estimate the average and peak number of transactions per second, and evaluate their average duration. Then, calculate the number of required concurrent connections as indicated in this example, and add a 5-10% margin. Then, monitor the back end, and adjust this value as required, to ensure that the back-end application can process all incoming requests.

When deploying adapters to a cluster, set the value of this attribute to the maximum required load divided by the number of cluster members.

For more information about how to size the back-end application, see the Scalability and Hardware Sizing document and the hardware calculator spreadsheet at https://developer.ibm.com/mobilefirstplatform/documentation/additional-documents/.


Parent topic: The adapter XML File