A complete solution for building and managing
Composable Clustered Infrastructure using GigaIO FabreX™ and NVIDIA Bright Cluster Manager
Build platform-independent compute clusters that scale in minutes to handle the most demanding AI and HPC workloads.
“Our research requires that we aggregate disparate computational elements, such as FPGAs, GPUs, x86 processors, and storage systems into highly usable and reconfigurable systems. GigaIO’s FabreX technology combined with NVIDIA Bright Cluster Manager makes it possible to dynamically bring these elements together in a very low-latency, high-performance interconnect while allowing for distinct, non-interfering workflows to co-exist on the same infrastructure.”
“Enterprises building AI and HPC computing infrastructure are seeking solutions that provide performance, flexibility, and efficiency. With native support for Bright Cluster Manager 9.2, GigaIO FabreX customers can compose and manage their compute systems to suit the needs of unique workloads from a single management interface.”
“Seamlessly integrating GigaIO’s FabreX and NVIDIA’s Bright Cluster Manager provides computing harmony, agility, and improved utilization of dynamic, growing environments. The performance of compute-intensive or data-intensive applications can be significantly improved via execution on the FabreX-BCM ecosystem.”
The Seamless GigaIO and NVIDIA Solution for
Rack-Scale Composable Computing
Both on-premise and in-cloud options for deploying workloads have significant limitations. An ideal solution would combine the flexibility of the cloud with the reliability, security, and cost-effectiveness of an on-premises configuration.
Users want a no-hassle solution that is easy to set up, manage, order, and install. A complete solution would handle everything from their application (or their container) on down. Users want to spend as little time as possible thinking about infrastructure, and they expect robust, enterprise-class capabilities in cluster management software, such as easy provisioning and updating, full access control, and monitoring.
GigaIO and NVIDIA have created a complete end-to-end solution for building reconfigurable infrastructure and made it easy to deploy and manage.
What is GigaIO FabreX?
GigaIO’s FabreXTM is a universal dynamic memory fabric that enables an entire server rack to be treated as a single compute resource, without proprietary architecture lock-in. No longer is the server the basic unit of compute. With FabreX, your entire rack is your server for true rack-scale computing.
Resources normally located inside of a server, including accelerators, storage, and even memory, can now be pooled in accelerator or storage enclosures, where FabreX makes them available to all of the servers in a rack.
These enclosures continue to communicate over native PCIe just as they would if they were plugged into the server motherboard. Every server connected to FabreX can also share its internal resources with every other server, enabling an entire rack of equipment to be treated like a collection of disaggregated puzzle pieces that can be re-assembled on the fly based on the specific requirements of each workload.
What’s New in NVIDIA Bright Cluster Manager 9.2?
NVIDIA Bright Cluster Manager is an enterprise-class software solution that simplifies building and managing HPC clusters from edge to core to cloud, transparently to the customer, by combining provisioning, monitoring, and management capabilities in a single tool.
Version 9.2 extends the goals of eliminating complexity and enabling flexibility by adding built-in support for composable infrastructure using GigaIO FabreX, where nodes can now be composed using NVIDIA Bright Cluster Manager Shell or BrightView.
How it Works
The GigaIO and NVIDIA solution combines NVIDIA Bright Cluster Manager’s ability to build and manage clusters with GigaIO FabreX’s ability to disaggregate and compose servers, accelerators, storage and memory together within a seamless memory fabric.Auto-scaling in Bright Cluster Manager creates a dynamic, multi-purpose infrastructure, and FabreX extends that agility to each hardware element in a rack by enabling the creation of composable GigaIO GigaPods and GigaClusters with cascaded and interlinked switches.
PCIe devices can be monitored and health-checked, as well as pooled and assigned to nodes within a cluster using Bright Cluster Manager. Clusters may have several different fabrics defined, and Bright Cluster Manager streamlines the fabric configuration process.
Support for FabreX in NVIDIA Bright Cluster Manager allows users to handle more workloads while maximizing resource utilization, minimizing cost, and managing everything from a single state-of-the art user interface. Cloud-like agility is now easier for on-prem infrastructure, allowing cloud bursting as needed within a single interface.
The NVIDIA Bright Cluster Manager and GigaIO FabreX solution provides a way to build easy-to-manage compute clusters that meet the needs of your most demanding workloads.
- Create “impossible serversTM”
- Deploy scale in minutes
- Maximize resource utilization
- Heterogeneous compute
- Easy to deploy and maintain
- Integrated build, management & monitoring
GigaIO’s Open Platform Ecosystem
The new integration is an example of GigaIO’s strategy to deliver an open platform that allows customers to access the benefits of composable infrastructure via the enterprise-class tools they already use.
GigaIO’s dedication to native integration with software tools like Bright Cluster Manager provides a best-in-class experience for customers who can continue to use their favorite tools without needing to alter their software stack. Native integration frees customers from having to rely on plug-ins with limited capabilities, learn new software, manage yet another pane of glass, and pay additional per-node license fees.