Novodisq is a New Zealand–based hardware and systems company focused on re-engineering data infrastructure to address the growing constraints of power, space, and data sovereignty in modern data centers. Founded in 2018, the company aims to become the backbone for sovereign and AI-ready data lakes by delivering ultra-dense, ultra-efficient storage and compute platforms designed for long-lived, data-heavy workloads. Novodisq positions its technology as a response to the rapid growth of global data—estimated at 20–30% annually—at a time when data-center power availability, cooling capacity, and physical space are increasingly limited.
The core problem Novodisq addresses is that most enterprise and AI data is neither hot nor archival, but “warm” data that must remain online, accessible, and retained for many years. This layer is traditionally served by power-hungry spinning disks or costly hyperscaler services, both of which scale poorly under today’s power and sovereignty constraints. As AI workloads grow, GPU clusters increasingly sit idle due to data ingestion bottlenecks, power shortages, and inefficient storage economics. Governments and enterprises are also demanding greater control over where data is stored and processed, driving interest in sovereign, on-premises infrastructure.
Novodisq’s solution is a modular, hardware-first architecture optimized for density, efficiency, and control. The flagship product, Novoblade™, is a 2U blade system that integrates high-density storage and compute in a single chassis. Each blade delivers up to roughly 576 TB, and a fully populated 2U system with 20 blades scales to about 11.5 PB while using up to 90–95% less power than traditional HDD- or flash-based storage systems. The design emphasizes watts-per-petabyte efficiency, enabling deployments in power-constrained data centers, regional facilities, or even edge locations.
A key differentiator is Novodisq’s vertically integrated hardware approach. After early proofs of concept using off-the-shelf components, the company chose to design its own SSDs, firmware, and system architecture to tightly control power consumption, cooling, and long-term reliability. The platform is built around low-power SoCs and FPGA-based acceleration, offloading functions such as RAID, checksumming, encryption, and data processing from CPUs. This enables high performance for write-once, read-sometimes data over a targeted 10-year hardware lifespan, with design trade-offs optimized for long-term retention rather than peak IOPS.
Alongside Novoblade, Novoforge™ serves as a development and pilot platform, allowing customers to test workloads, validate software stacks, and experiment with FPGA-accelerated data processing in a smaller, lab-friendly form factor. Use cases highlighted include genomics and pathology data, backup and restore staging, CCTV and NVR systems, Kubernetes and microservices clusters, and private cloud environments requiring strict data sovereignty. In several scenarios, Novodisq emphasizes the ability to ingest, process, and store data locally—reducing reliance on hyperscalers and improving time-to-recovery and operational resilience.
Novodisq is currently at an early commercial stage, engaging pilot customers and selling MVP hardware, with plans to layer in support and software services over time. Overall, the company positions itself as a dense, power-efficient alternative to legacy storage vendors and hyperscalers, enabling organizations to deploy AI-ready, sovereign data infrastructure in a world increasingly constrained by energy, space, and regulation.
0 commentaires:
Post a Comment