Load balancing back-end application services utilizing derivative-based cluster metrics

A computer readable data storage memory storing computer-executable instructions that, when executed by a computer system, cause the computer system to perform a computer-implemented method, the instructions comprising:

• instructions for receiving timestamped traffic data that characterizes amount of application service requests from one or more front-end servers of a social networking system to a server group of one or more back-end servers in the social networking system, and wherein the server group corresponds to a geographical tier in a server group hierarchy of application service back-end servers in the social networking system;
• instructions for receiving one or more timestamped metric measurements in a performance metric dimension and a performance metric threshold corresponding to the performance metric dimension and the geographical tier, wherein the timestamped metric measurements are for the server group;
• instructions for estimating a linear derivative between variable traffic data and variable performance metric in the performance metric dimension based on collected sample points respectively representing the received timestamped traffic data and the received timestamped metric measurements at different time points;
• instructions for computing, based on the estimated linear derivative, a threshold traffic capacity of the server group before the variable performance metric is estimated to exceed the performance metric threshold;
• and
• instructions for generating, based on the threshold traffic capacity, a routing table for access by a front-end server to facilitate selection of a back-end server from amongst the application service back-end servers, wherein the threshold traffic capacity determines how frequent that the front-end server selects a target back-end server outside of the server group as the back-end server destination, and wherein the application service back-end servers are potential traffic destinations in the routing table.

A computer system, comprising:

• non-transitory memory configured to store executable instructions;
• one or more processors configured by the executable instructions to:
• receive timestamped traffic data that characterizes amount of application service requests from one or more front-end servers of a social networking system to a server group of one or more back-end servers in the social networking system, and wherein the server group corresponds to a geographical tier in a server group hierarchy of application service back-end servers in the social networking system;
• receive one or more timestamped metric measurements in a performance metric dimension and a performance metric threshold corresponding to the performance metric dimension and the geographical tier, wherein the timestamped metric measurements are for the server group;
• collect and pair the timestamped traffic data and the timestamped metric measurements into sample points based on timestamps of the timestamped traffic data and the timestamped metric measurements;
• estimate a linear derivative between variable traffic data and variable performance metric in the performance metric dimension based on collected sample points respectively representing the received timestamped traffic data and the received timestamped metric measurements at different time points;
• compute, based on the estimated linear derivative, a threshold traffic capacity of the server group before the variable performance metric is estimated to exceed the performance metric threshold;
• and
• generate, based on the threshold traffic capacity, a routing table for access by a front-end server to facilitate selection of a back-end server from amongst the application service back-end servers, wherein the threshold traffic capacity determines how frequent that the front-end server selects a target back-end server outside of the server group as the back-end server destination.