Scheduling tasks at dynamically-allocated resource pools using share identifiers and fairness parameters

A computer-implemented method, comprising:

• acquiring, at a task scheduling service of a cloud provider network from a computing service of the cloud provider network, a resource pool of resources to execute a queue of tasks, wherein the resource pool includes dynamically allocated computing resources and an amount of resources present in the resource pool changes over successive scheduling iterations of the task scheduling service;
• conducting, at the task scheduling service, a plurality of scheduling iterations with respect to the queue of tasks and the resource pool, wherein individual tasks to be executed using the resource pool have respective share identifiers corresponding to respective portions of a capacity of the resource pool, wherein a particular scheduling iteration of the plurality of scheduling iterations comprises at least:
• identifying, by the task scheduling service and from a current task set comprising at least (a) one or more tasks present in the queue and (b) one or more currently-executing tasks which were in the queue prior to their execution, a set of currently-in-use share identifiers;
• determining, by the task scheduling service and using at least a fairness-based reservation parameter, a first subset of the resource pool to be reserved for assignment to tasks with share identifiers which are not members of the set of currently-in-use share identifiers;
• assigning, by the task scheduling service and to one or more tasks selected from the queue based at least in part on their respective share identifiers, one or more resources from a second subset of the resource pool, wherein the second subset is determined at least in part from the first subset;
• and
• causing, by the task scheduling service, the computing service to execute the one or more tasks selected from the queue on the one or more resources of the second subset according to the assigning, wherein use of the at least a fairness-based reservation parameter in the determining of the first subset of the resource pool to be reserved for assignment to tasks with share identifiers which are not members of the set of currently-in-use share identifiers reduces a likelihood that the one or more tasks selected from the queue are starved of resources.

A non-transitory computer-accessible storage medium storing program instructions that when executed on a processor cause the processor to:

• acquire, at a task scheduling service of a cloud provider network from a computing service of the cloud provider network, a resource pool of resources to execute a queue of tasks, wherein the resource pool includes dynamically allocated computing resources and an amount of resources in the resource pool changes over successive scheduling iterations of the task scheduling service;
• conduct, at the task scheduling service, a plurality of scheduling iterations with respect to the queue of tasks and the resource pool, wherein individual tasks to be executed using the resource pool have respective share identifiers corresponding to respective portions of a capacity of the resource pool, wherein a particular scheduling iteration of the plurality of scheduling iterations comprises at least:
• identifying, by the task scheduling service and from a current task set comprising at least (a) one or more tasks present in the queue and (b) one or more currently-executing tasks which were in the queue prior to their execution, a set of currently-in-use share identifiers;
• determining, by the task scheduling service and using at least a fairness-based reservation parameter, a first subset of the resource pool to be reserved for assignment to tasks with share identifiers which are not members of the set of currently-in-use share identifiers;
• assigning, by the task scheduling service and to one or more tasks selected from the queue based at least in part on their respective share identifiers, one or more resources from a second subset of the resource pool, wherein the second subset is determined at least in part from the first subset;
• and
• causing, by the task scheduling service, the computing service to execute the one or more tasks selected from the queue on the one or more resources of the second subset according to the assigning, wherein use of the at least the fairness-based reservation parameter in the determining of the first subset of the resource pool to be reserved reduces a likelihood that the one or more tasks selected from the queue are starved of resources.