In this article we look at several introductions of digital storage related products at the 2023 Supercomputing Conference. Digital storage and memory are critical components in high performance computing, including AI. In particular, introductions by DDN and Arcitecta and WDC for software defined scale-out storage and high-performance storage applications.
DDN announced its Infinia software-defined storage (SDS) platform targeted for AI and cloud applications. The figure below shows attributes of this SDS system.
The core data plane is based upon flash memory for all the storage, not just primary storage, including the use of QLC flash for higher latency storage, scaling up to 100’s of PB of storage. Writes are bundled in cache to reduce wear on the QLC flash. It is meant to fit within existing customer orchestration tools and is hardware agnostic and supports multi-tenancy. The Infinia is meant to support applications that don’t need the performance and scale of the company’s ESAScaler storage systems.
The Infinia system is object based at present but longer term it is intended to support containers and files as well as objects, with each service having its own container. DDN will be showing AMD-based 1U servers as reference architectures with 12 hot-swapable 2.5” U.2 NVMe drive bays and up to four 200G network ports, with a minimum of 6 or these used to create a cluster acting as an S3 object store.
In 2024 DDN says they will have a parallel filesystem designed for AI and cloud and support advanced database and data movement.
Australia-based data management software company, Arcitecta announced its MediaFlux Universal Data System, which combined data management, data orchestration, multi-protocol access and storage in a single platform. It can manage data both on-premises and in the cloud and includes globally distributed access for 100’s of billions of files. The company says that it can do this at an order of magnitude less cost than legacy cluster file systems.
Arcitecta has been built on top of third-party file and object storage systems from other vendors, as shown below. An example is its use with NetApp E-Series block storage although their biggest market is with Isilon storage systems.
Cluster file system storage runs at $100-$1,000 per TB but Arcitecta licensing is per user rather than by capacity. They told me a price as low as $100 per user, irrespective of storage capacity used. There can still be a maintenance fee.
Western Digital was showing the company’s OpenFlex Data24 3200 NVMe-oF (NVMe over fabric) storage platform. This system can create a storage pool that can be shared by up to six hosts without a switch and supports remote direct memory access (RDMA) over converged Ethernet (RoCE) and is available in a 2U 24-bay platform with storage capacities up to 368TB using the company’s Ultrastar DC SN655 dual-port PCIe Gen 4.0 SSDs.
WDC was also showing its hybrid storage JBOD Ultrastar Data102 and Data60 platforms to support disaggregated storage and software-defined storage (SDS). This comes in dual-port SAS or single-port SATA configurations. The Data102 has storage capacities up to 2.65PB and the Data60 has up to 1.56TB in a 4U enclosure that includes IsoVibe and ArticFlow technologies for improved performance and reliability. The Data102 and Data60 capacity numbers assume using 26TB SMR HDDs.
WDC was also showing a GPUDirect storage proof of concept combining the company’s RaidFlex technology with Ingrasys ES2100 with integrated NVIDIA Spectrum Ethernet switches as well as NVIDIA’s GPUs, Magnum IO GPUDirect storage, BlueField DPUs and ConnectX SmartNICs. The proof-of-concept demonstration can provide 25GB/s bandwidth for a single NVIDIA A100 Tensor Core GPU and over 100GB/s for four NVIDIA A100 GPUs.
At SC23 Arcitecta and DDN introduced software defined storage solutions for AI and cloud applications. WDC was also showing SDS, its OpenFlex NVMe storage and GPUDirect storage.