FabreXTM AI Memory Fabric Platform
The enormous increase in data is driving fundamental change to data center architecture
The remarkable increase in the amount of data being collected, and that must be analyzed and stored, is driving the rapid adoption of advanced data analytics and Artificial Intelligence (AI). Both are challenging the fundamental architectures of today’s data centers, a scale of disruption not seen since the 1990s.
AI, and the associated Machine Learning (ML) and Deep Learning (DL) applications are creating new challenges for compute and storage clusters. Faster and larger storage arrays, and a proliferation of specialized compute accelerators, like GPUs, FPGAs and custom ASICs, are creating bottlenecks and configuration\ problems for the interconnect systems, as the traditional networks were never designed to handle the performance requirements of these workflows and devices.
Further, the rapid pace of change in acceleration technology and AI software fuels the necessity for flexible and easy to upgrade architectures, capable of incorporating new technology without demanding forklift upgrades to expensive equipment. This means disaggregating the elements of the traditional server into separate components that can be easily shared. But, in order to effectively disaggregate storage and accelerators, the interconnects must support both an exceptionally low latency AND high bandwidth.
A related challenge is the low utilization (15% on average) of these expensive new accelerators, stranded in the server sheet metal in siloed configurations. The result for High Performance Computing, Enterprise, Cloud and Edge data centers is the need for both scale-up and then scale-out resources across the cluster with a technology that will grow in both directions.
FabreX is the next generation, memory-centric fabric for a changing compute world. Effortlessly connect new memory / storage products, the multitude of new accelerators and your choice of processors either directly attached or via server configs like NVMe-oF.
FabreX delivers the industry’s lowest latency AND the highest effective bandwidth. Latency from system memory of one server to system memory of any other is less than 200ns – true PCIe performance across the entire cluster. The FabreX Gen4 implementation scales up to 512Gbits/sec bandwidth.
FabreX can compose an unprecedented variety of resources, such as accelerators including GPUs, DPUs, TPUs, FPGAs and SoCs or storage devices, such as NVMe, PCIe native storage, and other I/O resources connected to compute nodes. In addition to device-to-node, FabreX is unique in enabling node-to-node and device-to-device communication all across the same high performance PCIe memory fabric. FabreX can span multiple servers and multiple racks to scale up single-server systems and scale out multi-server systems, all unified via the FabreX software.
Featuring 100% PCI-SIG compliance, the FabreX switch can integrate heterogenous computing, storage, and accelerators into one system-area cluster fabric, so you can do more with less. Patented GigaIO technology strips away unnecessary conversion, software layers and overheads that add latency to legacy interconnects.
FabreX is built on, and 100% compliant with, the industry’s most widely adopted standard, PCIe, insuring low risk, easy integration and long life.
Further, the FabreX operating system easily integrates third-party applications with its open-source design, including the DMTF open-source Redfish APIs to provide unprecedented integration with a range of third-party applications for fabric automation, orchestration, resource allocation and job management.
The Solution: GigaIO FabreXTM
GigaIO FabreX is a fundamentally new composable architecture that integrates computing, storage and other communication I/O into a single-system cluster fabric, using industry standard PCI Express (PCIe) technology. PCIe is used internally to connect every server on the market today. GigaIO enables true server-to-server communication across PCIe and makes cluster scale computer possible, with direct memory access by an individual server to system memories of all other servers in the cluster fabric, for the industry’s first in-memory network.
This new architecture enables a unified, software-defined, and memory-centric composable infrastructure.
With its exceptional low latency and high bandwidth, FabreX makes the disaggregation of all server elements possible. Standard Redfish® APIs enable integrations with leading third-party orchestration and composition software for a true software-defined infrastructure that dynamically assigns resources to match changing workloads. Gone are the days of having to make massive and costly changes to large numbers of servers to incorporate new technology.
Upgrade or add compute, storage and application accelerators at the component level that plug-n-play with your environment. Every major subsystem can now operate on its own upgrade cycle. And the total cost of the system is optimized as FabreX drives much higher utilization of the resources.
FabreX is built to be an open platform and offers robust, industry standard APIs at all levels of the software making the integration of FabreX into your existing data center management systems easy and risk free.
FabreX is 100% compliant with the industry’s leading standard, PCIe, insuring high levels of support and compatibility from an enormous selection of vendors and technologies. Every new storage, acceleration and compute technology fully support the latest PCIe standards ensuring full access to the latest and most capable technologies.
And as IT experts begin to use new architectures for moving compute to the data at the edge of your networks, FabreX is a natural choice, offering the greatest depth of support in storage, compute and acceleration technology, the highest density, lowest power and best performance.
The FabreXTM Building Blocks
FabreX Fabric Manager
The FabreX Fabric Manager (FM) is the central building block and the software engine that drives the performance and dynamic composability of GigaIO’s software-defined hardware. This Linux-based, resource-efficient software layers onto FabreX hardware for easy-to-use composing of computing clusters on-the-fly.
The FabreX FM is designed to be an open environment with robust APIs to ease integration and encourage third party developers to enhance the usability, flexibility and performance of the fabric. You continue to use the tools you know and trust – with an entire new layer of capability.
The FabreX Switch
A classic Top-Of-Rack network switch, the FabreX Switch communicates with FabreX server drivers to identify and coordinate resources required by the servers.
FabreX Network Adapter Card
The FabreX Network Adapter card is the high-performance, cabled interface to cluster subsystems across FabreX. The card includes both server and target (for PCIe I/O) modes and is FPGA-powered for configuration flexibility. Applications can access remote PCIe devices as if they were attached to the local system.
The FabreX Network Adapter Card is designed for maximum cable length at full PCI Express Gen 3.0 or Gen 4.0. Copper cables can connect up to 3 meters or use active optic cables to extend the distance up to 100 meters.
The flexibility of the FabreX platform breaks the constraints of old architectures, opening up possibilities for new configurations that maximize utilization of all elements within your High Performance Computing, Enterprise, Cloud or Edge data centers. At the same time, FabreX allows for far greater utilization of these next generation technologies, keeping both your capital expenditures and operating costs down. Cloud flexibility at a fraction of the cost.
