Paper Title: Memory-Converged Infrastructure Based on Persistent Memory

Paper Abstract: Memory-Converged Infrastructure (MCI) is a new architecture that combines the benefits of persistent memory, hyperconverged architecture, and container orchestrations. Leveraging new persistent memory implementations based on 3D XPoint technology, MCI delivers both bigger memory and faster storage with few or no application rewrites. It targets data-intensive workloads such as machine learning and big data analytics and promises to remove two major bottlenecks, namely memory capacity and storage I/O performance. A recent MCI implementation called Distributed Memory Object converges compute and storage onto a single cluster of hyperconverged nodes with persistent memory inside. Data-centric applications can be easily deployed on the platform using Kubernetes-based orchestration. The implementation has been tested in large customer environments, and produced clear improvement over the existing infrastructure.

Paper Author: Charles Fan, Co-founder & CEO, MemVerge

Author Bio: Charles Fan is co-founder and CEO of MemVerge, an early-stage startup building Memory-Converged Data Infrastructure software on top of the new persistent memory technologies. Prior to MemVerge, Charles was CTO of Cheetah Mobile, leading its technology teams in AI, big data and cloud. Before Cheetah, Charles was a SVP/GM at VMware, responsible for VMware’s storage and availability business unit. He led the teams that created the HCI market leader Virtual SAN. Before VMware, Charles worked at EMC and was the founder of EMC China R&D Center. Charles joined EMC via the acquisition of Rainfinity. He was a co-founder and CTO of Rainfinity and was responsible for its file virtualization and high availability products. Charles received his Ph.D. and M.S. in Electrical Engineering from the California Institute of Technology, and his B.E. in Electrical Engineering from the Cooper Union.