Dynamic latency shaping to control server load

A method implemented at a performance monitoring system in communication with a storage host that hosts block storage volumes, wherein the storage host is in communication with a compute host hosting a particular compute instance of compute instances, and wherein the compute host facilitates submission of input/output (I/O) operations to the block storage volumes by the particular compute instance, the method comprising:

• receiving load information for the storage host;
• determining that the load information satisfies a load threshold;
• selecting the particular compute instance on which to modify a performance profile to modify a latency of at least a subset of the I/O operations of the one or more of the compute instances;
• modifying the performance profile of the particular compute instance to modify the latency of the subset of I/O operations, and wherein the subset of I/O operations is selected to modify load of the storage host according to the load threshold;
• and
• sending the modified performance profile to the storage host, wherein the storage host upon receiving the modified performance profile selectively adds synthetic latency to the subset of I/O operations according to the modified performance profile.

One or more non-transitory computer-readable media comprising instructions executable by a performance monitoring system in communication with a storage host hosting block storage volumes, wherein the storage host is in communication with a compute host hosting a particular compute instance of compute instances, and wherein the compute host facilitates submission of input/output (I/O) operations to the block storage volumes by the particular compute instance, and wherein the instructions, when executed, cause the performance monitoring system to:

• receive load information for the storage host;
• determine that the load information satisfies a load threshold;
• select the particular compute instance on which to modify a performance profile to modify a latency of at least a subset of the I/O operations of the particular compute instance;
• modify the performance profile of the particular compute instance to modify the latency of the subset of I/O operations, and wherein the subset of I/O operations is selected to modify load of the storage host according to the load threshold;
• and
• send the modified performance profile to at least one device of the compute host or the storage host, wherein the at least one device is configured to selectively add synthetic latency to the subset of I/O operations according to the modified performance profile.