A fleet of hundreds of EC2 instances each has two 1 GiB gp2 EBS volumes. A critical workload is saturating the EBS IOPS. Company rules forbid changing instance types or volume types without lengthy testing. What is the fastest way to improve EBS I/O performance without altering instance or volume types?

Increasing the provisioned size of gp2 EBS volumes directly increases their baseline IOPS and burst performance. For gp2 volumes, IOPS scale linearly with size, up to a maximum of 16,000 IOPS for volumes 5,334 GiB or larger. Even a small increase from 1 GiB will provide a significant boost, as 1 GiB gp2 volumes have a baseline of only 3 IOPS. Adding extra ENIs would improve network throughput, not EBS I/O performance. Transfer Acceleration is for S3, not EBS. Cluster placement groups improve network latency and throughput between instances, not EBS performance.