Database query distribution

A method, comprising:

• receiving, from a client, a query targeting a data set at a query placement system for a managed database service that automatically manages a size of a pool of processing clusters used to perform queries to the data set;
• identifying the pool of processing clusters as associated with the client, respectively capable of processing the query;
• randomly selecting a first subset of two or more processing clusters from the pool of processing clusters;
• sending the query to a first selected processing cluster in the first subset of two or more processing clusters, wherein the first selected processing cluster is selected by the query placement system according to utilization data for the pool of processing clusters maintained at the query placement system;
• receiving a response from the first selected processing cluster of the first subset of two or more processing clusters indicating that the first selected processing cluster does not accept the query for execution according to an evaluation of local utilization data maintained at the first selected processing cluster, wherein the response includes an estimated utilization of computing resources to execute the query at the first selected processing cluster that was considered in combination with the local utilization data with respect to a utilization threshold, wherein the consideration of the estimated utilization in combination with the local utilization data with respect to the utilization threshold determines that the first selected processing cluster cannot accept the query for execution;
• before making a determination to modify the size of the pool of processing clusters according to the estimated utilization received from the first selected processing cluster:
• selecting a second subset of two or more processing clusters of the pool of processing clusters that is different from the first subset of two or more processing clusters;
• using the estimated utilization of computing resources to execute the query determined by and received from the first selected processing cluster to select, from the second subset of two or more processing clusters, a second processing cluster based, at least in part, on the utilization data for the pool of processing clusters maintained at the query placement system;
• and
• sending the query to the second processing cluster.

One or more non-transitory, computer-readable storage media storing instructions that, when executed on or across one or more processors, cause the one or more processors to implement a query placement system for a managed database service, wherein the query placement system implements:

• receiving, from a client, a query targeting a data set at the placement system, wherein the database management service automatically manages a size of pool of processing clusters used to perform queries to the data set;
• identifying the pool of processing clusters as associated with the client, respectively capable of processing the query;
• randomly selecting a first subset of two or more processing clusters from the pool of processing clusters;
• sending the query to a first selected processing cluster in the first subset of two or more processing clusters, wherein the first selected processing cluster is selected by the query placement system according to utilization data for the pool of processing clusters maintained at the placement system;
• receiving a response from the first selected processing cluster of the first subset of two or more processing clusters indicating that the first selected processing cluster does not accept the query for execution according to an evaluation of local utilization data maintained at the first selected processing cluster, wherein the response includes an estimated utilization of computing resources to execute the query at the first selected processing cluster that was considered in combination with the local utilization data with respect to a utilization threshold, wherein the consideration of the estimated utilization in combination with the local utilization data with respect to the utilization threshold determines that the first selected processing cluster cannot accept the query for execution;
• before making a determination to modify the size of the pool of processing clusters according to the estimated utilization received from the first selected processing cluster:
• selecting a second subset of two or more processing clusters of the pool of processing clusters that is different from the first subset of two or more processing clusters;
• using the estimated utilization of computing resources to execute the query determined by and received from the first selected processing cluster to select, from the second subset of two or more processing clusters, a second processing cluster based, at least in part, on the utilization data for the pool of computing resources maintained at the placement system;
• and
• sending the query to the second processing cluster.