Abstract: Embodiments of the present invention provide concepts for handling incoming transaction requests to a transaction server based on a historical record of the server resources required to execute similar transactions in the past and the current capacity of the transaction server. Incoming transaction requests may be classified as heavy weight transactions or light weight transactions and the handling of the transaction requests may be based on said classification.

Abstract: Controlling a server latency response is presented. One example comprises monitoring, at a server, a number of requests from a client that cause a failure response. The method then comprises, responsive to the number of requests meeting a predetermined requirement, delaying processing of the requests from the client by a predetermined delay time.

Abstract: Controlling a server latency response is presented. One example comprises monitoring, at a server, a number of requests from a client that cause a failure response. The method then comprises, responsive to the number of requests meeting a predetermined requirement, delaying processing of the requests from the client by a predetermined delay time.

Abstract: Embodiments for resolving outstanding affinities between a server cluster member and its clients are presented. Outstanding affinities between the server and the client may be identified in response to the server restarting. A reconnection request may then be generated and communicated to the client. Responsive to a first server of a server cluster restarting following a failure of a connection between a client and the first server, identifying, by an affinity analysis component, the outstanding affinity between the first server and the client. Responsive to identifying the outstanding affinity between the first server and the client, communicating, by a reconnection component, a reconnection request for reestablishing the connection between the first server and the client.

Abstract: Avoiding failure repetition in data processing includes storing a sequence of circumstances leading up to a previous failure, monitoring circumstances in a current process, matching a sequence of circumstances in the current process to a stored sequence of circumstances, and applying rules to determine if the current process should proceed.