A distributed system of data centers is configured to ensure that all data operations involving specific accounts are performed within a given time frame at only one data center. Once a data center has begun accessing data that is associated with an account, all requests to access or manipulate data for that account are redirected to that data center for a period of time. Thus, the data center becomes not only the “owner” of all data associated with the account for that period of time, but also the servicer of all requests associated with the account. The data center then asynchronously replicates changes to the data associated with the account to the other data centers. Since none of the other data centers are permitted to change the data associated with the account, conflicts in the replicated data are minimized.