NVIDIA has acquired Tel-Aviv based high-performance software-defined storage company Excelero. The acquisition is in line with the chip giant’s plan to expand its capabilities in HPC and AI. The terms of the deal have not been disclosed.
Excelero has raised USD 35 million across four funding rounds from investors, including Western Digital’s venture arm, Battery Ventures, Qualcomm Ventures etc.
Excelero enables customers to build distributed, high-performance Server SAN with standard hardware for applications at any scale. The acquisition will support block storage in NVIDIA’s enterprise software stack. Additionally, block storage is critical for the DOCA software framework that runs on DPUs. “We’re thrilled to apply our expertise in block storage to NVIDIA’s world-class AI and HPC platforms,” said Yaniv Romem, CEO and co-founder of Excelero.
NVMesh
Excelero’s NVMesh is a software that manages and secures virtual arrays of NVMe flash drives as block storage available across public and private clouds. As per Excelero, ephemeral NVMe storage provides fast performance but insufficient data protection or data sharing. NVMesh software-defined storage deploys distributed data protection over public cloud instances with ephemeral NVMe drives, creating a unique, virtual, high-performance and low-latency storage pool. NVMesh supports instances, both virtualised and containerised, in public cloud environments that feature NVMe drives.
The solution features an intelligent management layer that abstracts underlying hardware with CPU offload, creates logical volumes with redundancy, and provides centralised management and monitoring. Customers benefit from the performance of local flash, with the convenience of centralised storage and the cost savings of standard hardware. The solution has been deployed for Industrial IoT services, machine learning applications and simulation visualisation.
Image: Excelero
NVMesh can be deployed on a range of supported server platforms and works well with hardware from different vendors at various capacities across generations.
Excelero and NVIDIA go way back
Israel has been a key location for NVIDIA. Last year, the chip giant announced that it would expand its Israel operations and hire 600 new personnel, including electrical engineers, software engineers, computer science engineers, chip designers, architects (hardware and software), and quality control specialists.
In 2019, NVIDIA bought Israel’s Mellanox Technologies (communications infrastructure for data servers supplier) for USD 7 billion. The goal was to enable customers to achieve higher performance with better utilisation of computing resources and lower operating costs.
Excelero has worked with major cloud service providers. The firm’s software is known for high throughput, low latency and support for Kubernetes containers. The company has been an NVIDIA partner from early days.
NVIDIA developed the Magnum IO software suite with the help of Excelero. AI researchers use the suite to process vast amounts of data quickly. Magnum IO GPUDirect Storage and Excelero’s Remote Direct Drive Access (RDDA) provide a direct GPU-to-drive path for the most latency-sensitive applications, the company said. NVIDIA’s Quantum-2 Infiniband networking platform uses Excelero’s software.
Enhancing HPC capabilities
The cloud High Performance Computing (HPC) market was valued at USD 4.50 billion in 2020 and will reach USD 11.54 billion by 2026, at a CAGR of 16.68%, as per a report by Mordor Intelligence.
NVIDIA understands the potential of HPC and AI, and has been acquiring companies to push growth. Excelero is NVIDIA’s second HPC centric acquisition in 2022. The chip giant acquired Bright Computing, a software provider for HPC, just months back to push the latter to more markets. Bright’s software will consolidate NVIDIA DGX and data centre businesses.
Founded in 2009 in Amsterdam, Bright Computing has big clients such as Boeing, NASA, Johns Hopkins University and Siemens. Its software can run at the edge, data centre and across multiple public or hybrid clouds. It supports Arm and x86 CPUs, NVIDIA GPUs, and Kubernetes containers.
Many companies leverage NVIDIA Omniverse to build physically accurate 3D simulations and digital twins for multiple functions such as product design automation, drug discovery, etc.