Disaggregated Storage using OPI and Marvell Octeon DPUs

A prominent trend in disaggregated storage is the use of Non-Volatile Memory Express over Fabric (NVME-oF) and in particular, NVMe over TCP to connect storage devices over a network. But there is no straightforward to provide this storage to Virtual machines and containers. The hypervisor will still need to emulate virtio-blk or virtio-scsi kind of emulated interfaces to expose this storage which involves usage of hypervisor processor cycles. Marvell Octeon DPUs provide a local NVMe to remote NVMeOF storage offload, where the NVMe devices can appear as local PCIe controllers to the VMs and containers but can reach to namespaces provided over NVMe-OF TCP target storage with the help of the DPU without the involvement of the host. As on-prem, hybrid and bare metal tenancy become more prevalent, most hyperscalers and datacenter operators are looking at DPUs and IPUs to provide streamlined and fine-grained control to their infrastructure. The DPU is designed to be a host-independent infrastructure endpoint that not only exposes the device’s storage services to the host but also at the same time securely exposes the device’s capabilities and provisioning control to the broader infrastructure and orchestration mechanisms. Marvell has collaborated with OPI(https://github.com/opiproject) to this end to expose the Octeon DPUs’ NVMe based storage Infrastructure offload capabilities. This talk will explore the details and capabilities of this solution.