If a distributed shared storage system is used to provide high bandwidth real-time media data that is shared by a large number of users, several complexities arise. In such an application, the high bandwidth real-time media data is distributed among multiple storage devices or servers. Multiple client applications or machines may access the data. To access the data, clients and servers communicate among themselves using short control messages that contain queries and responses, and using longer data messages that contain media data. To reduce this variability in latency, short control messages are posted for transmission immediately, whereas long messages are posted for transmission in a controlled fashion. For example, the number of long messages that may be outstanding at any one time may be limited. Thus, a long message is not posted to the transmit queue if the maximum number of long messages has been reached. A long message may be posted when transmission of an outstanding long message has been completed. In this way, the latency for any short message is limited by the amount of time it takes to send a long message and the number of long messages that are permitted to be outstanding at any one time.