Query admission control for online data systems based on response time objectives

A system comprising:

• at least one memory device;
• and
• a processing device, operatively coupled with the at least one memory device, to:
• track processing time metrics of multiple server queries;
• estimate a processing time per query type using the tracked processing time metrics;
• estimate a current queue wait time using a sum of products, each of the products being, for each query type, a result of multiplying a number of queries of that query type currently in the queue and the estimated processing time for that query type;
• receive a current server query from a client, the server query including a current query type;
• map the current query type to an estimated processing time determined using the tracked processing time metrics;
• determine an estimated response time using a sum of the current queue wait time and the estimated processing time for the current query type;
• reject the server query from being added to the queue in response to determining the estimated response time does not satisfy a service level objective;
• and
• send an error message to the client indicating the rejection of the server query.

An apparatus comprising:

• means for tracking processing time metrics of multiple server queries;
• means for estimating a processing time per query type using the tracked processing time metrics;
• means for estimating a current queue wait time based on a number of queries currently in the queue and the estimated processing times of query types for each of the queries currently in the queue;
• means for receiving a current server query from a client, the server query including a current query type;
• means for mapping the current query type to a first percentile estimated processing time determined using the tracked processing time metrics;
• means for determining a first estimated response time using the current queue wait time and the first percentile estimated processing time for the current query type;
• means for mapping the current query type to a second percentile estimated processing time determined using the tracked processing time metrics;
• means for determining a second estimated response time using the current queue wait time and the second percentile estimated processing time for the current query type;
• means for rejecting the server query from being added to the queue in response to determining the first or second estimated response time does not satisfy a service level objective;
• and
• means for sending an error message to the client indicating the rejection of the server query.