Creditless performance bursting in multi-tenant environments

A computer-implemented method comprising:

• executing, by a host computer system of a cloud provider network, a first burstable performance compute instance having an associated first baseline compute performance level, wherein the first burstable performance compute instance is able to use at least a portion of a compute capacity headroom of the host computer system to exceed the first baseline compute performance level;
• determining a total baseline capacity requirement associated with the host computer system, the total baseline capacity requirement to include the first baseline compute performance level;
• determining a compute resource utilization of a plurality of burstable performance compute instances hosted by the host computer system, wherein the plurality of burstable performance compute instances includes the first burstable performance compute instance;
• determining that a difference between the compute resource utilization and the total baseline capacity requirement is below a threshold;
• and
• updating a status identifier associated with a slot of a plurality of slots associated with the host computer system, each slot of the plurality of slots associated with a portion of a total compute capacity of the host computer system, the status identifier constraining the slot to prevent or limit placement of an instance in the slot to reserve the portion of the total compute capacity of the host computer system associated with the slot as additional headroom for other instances already hosted on the host computer system, wherein the reserved portion of compute capacity contributes to the compute capacity headroom.

A system comprising:

• a host computer system of a cloud provider network, the host computer system including instructions that upon execution cause the host computer system to:
• execute a first burstable performance compute instance having an associated first baseline compute performance level, wherein the first burstable performance compute instance is able to use at least a portion of a compute capacity headroom of the host computer system to exceed the first baseline compute performance level;
• a first one or more electronic devices to implement a service in the cloud provider network, the service including instructions that upon execution cause the service to:
• determine a total baseline capacity requirement associated with the host computer system, the total baseline capacity requirement to include the first baseline compute performance level;
• determine a compute resource utilization of a plurality of burstable performance compute instances hosted by the host computer system, wherein the plurality of burstable performance compute instances includes the first burstable performance compute instance;
• determine that a difference between the compute resource utilization and the total baseline capacity requirement is below a threshold;
• and
• update a status identifier associated with a slot of a plurality of slots associated with the host computer system, each slot of the plurality of slots associated with a portion of a total compute capacity of the host computer system, the status identifier constraining the slot to prevent or limit placement of an instance in the slot to reserve the portion of the total compute capacity of the host computer system associated with the slot as additional headroom for other instances already hosted on the host computer system, wherein the reserved portion of compute capacity contributes to the compute capacity headroom.