A SysOps administrator manages a self-managed relational database on a single EC2 instance. The database workload is characterized by small, random I/O operations and requires sub-millisecond latency with a sustained 32,000 IOPS. The application team also demands 99.999% volume durability for the transaction logs. The current gp3 volume is hitting its provisioned IOPS ceiling and experiencing queue depth spikes. Which EBS volume type should the administrator choose to meet these requirements?