When MQ first delivered shard queues, one significant benefit was the 'pull' model of message processing, the queue managers that were most efficient processed the most messages. The CICS regions with the lightest workload or on the fastest hardware got most of the work. In some circumstances, the ‘pull’ model is not the best model for the overall environment. When there are asymmetric sysplex environments there can be a significant workload skewing towards more powerful processors or towards LPARs getting better response time from a coupling facility and this can really impact the workload balance (and the pricing!). This session will outline the symptoms and patterns of workload imbalance we have seen and offer ways to help control the flow of the message driven CICS processing.