Adjusting dynamically scalable instance hosting based on compute resource usage

A system comprising:

• a host computer system coupled to a provider network, the host computer system to implement a monitoring agent and to host a plurality of instances;
• and
• a set of hardware virtualization services implemented by a plurality of computer systems coupled to the provider network, the set of hardware virtualization services including instructions that upon execution cause the plurality of computer systems to:
• receive a request to launch a dynamically scalable instance, the dynamically scalable instance to have a baseline performance level associated with a processor, a memory, and a network interface and a maximum performance level associated with the processor, the memory, and the network interface;
• select, based in part on the request, the host computer system to host the dynamically scalable instance from a plurality of host computer systems coupled to the provider network;
• cause the host computer system to launch the dynamically scalable instance, the dynamically scalable instance being one of the plurality of instances hosted by the host computer system;
• obtain time series data from the monitoring agent implemented by the host computer system, the time series data indicating amounts of processing utilization, memory utilization, and network utilization of the host computer system by each of the plurality of instances;
• and
• in response to a determination that a first resource usage value of the host computer system satisfies a first threshold, cause a migration of the dynamically scalable instance to a second host computer system of the plurality of host computer systems, the second host computer system having unused computing resources to allow the dynamically scalable instance to operate up to the maximum performance level associated with at least one of the processor, the memory, and the network interface, wherein the first resource usage value is based on the time series data corresponding to a first period of time.

A non-transitory computer-readable storage medium storing instructions which when executed by one or more processors of a computing device cause the computing device to perform operations comprising:

• receiving a request to launch a dynamically scalable instance, the dynamically scalable instance to have a baseline performance level associated with a processor, a memory, and a network interface and a maximum performance level associated with the processor, the memory, and the network interface;
• selecting, based in part on the request, a first computer system to host the dynamically scalable instance from a plurality of computer systems of a provider network;
• causing the first computer system to launch the dynamically scalable instance, the dynamically scalable instance being one of a plurality of instances hosted by the first computer system;
• obtaining time series data from a monitoring agent implemented by the first computer system, the time series data indicating amounts of processing utilization, memory utilization, and network utilization of the first computer system by each of the plurality of instances;
• and
• in response to a determination that a first resource usage value of the first computer system satisfies a first threshold, causing a migration of the dynamically scalable instance to a second computer system of the plurality of computer systems, the second computer system having unused computing resources to allow the dynamically scalable instance to operate up to the maximum performance level associated with at least one of the processor, the memory, and the network interface, wherein the first resource usage value is based on the time series data corresponding to a first period of time.