Abstract: A method of servicing requests for information at a server includes receiving a request for information from a client device, and determining a number of attempts of the request made by the client device. Both the client device and the server may count the number of attempts for each request. When the number of attempts is greater than a threshold, the method includes servicing the request and sending the information to the client device. When the number of attempts is not greater than the threshold, the method includes sending a service unavailable message to the client device. A retry-after delay may be calculated by the server according to a predicted time for the server to exit an overloaded state and included in the service unavailable message. The predicted time may take into account both requests currently being serviced and requests that have been scheduled to be serviced in the future.