HPE: AI at Hyperscale – How to go faster with a smaller footprint

Teresa Melvin, an AI-driven Big Data Solutions Architect at HPE, oversees the open-source solutions R&D lab in Fort Collins, Colorado. To meet the demands of persisting millions of I/O events per second, her team relies on Aerospike running on Intel persistent memory (pmem) DIMMs, achieving 280,000 sustained read/write operations per second—far surpassing other technologies. Aerospike’s ability to handle a 1 to 10 write-read ratio and its seamless integration into a Kubernetes dataflow pipeline enables real-time machine learning and flexible deployment across various environments. Additionally, Aerospike’s support for warm database restarts ensures minimal downtime during maintenance.