A method of detecting latency in message delivery in a communication network is disclosed. One or more data centers receives messages that are to be published via the communication network. The system receives queries from subscribers, processes the queries and transmits messages to the subscribers in response to the queries. The subscribers acknowledge the messages. The system uses its own processing time and the acknowledgement time to assess latency in message delivery. The system also may use the information to determine which data center or geographic location is experiencing relatively higher latency than other data centers or locations.