Break Infrastructure Constraints with a Dynamically Configurable Compute Environment
A Turnkey, Composable Appliance for Higher Education
Tired of having your critical resources imprisoned by current server configurations? Let the GigaIO Composability Appliance, University Edition, Powered by AMD, liberate them by disaggregating them from server sheet metal. Workloads for both classroom and research settings can now get exactly the resources they need.
Eliminate Constrained Access to Compute Resources
The GigaIO Composability Appliance, University Edition, creates a cloud-like environment where critical resources are universally accessible across departments at a fraction of cloud TCO.
Provide Students with The Latest Tech Every Single Day
You want to teach, and students want to learn. GigaIO and AMD can deliver access to the same technologies they’ll be using out in the real world. Preparing the next generation for the future is your charter — delivering access to next-generation technologies is ours.
Institutions of Higher Learning Are Under Constant Pressure
You need to educate future innovators, conduct bleeding-edge research, and do it all on a limited budget. Our engineered solution increases utilization of existing compute and GPU resources, and delivers broader access to limited resources via dynamically shared allocation. It levels the playing field by democratizing access for those laboring in the classroom as well as the lab.
To Become Proficient in Something You Need to Spend 10,000 Hours Doing it
Empower your students to become experts by giving them more access to the tools they’ll be using once they graduate. Enable them to research innovative concepts, and turn out better prepared professionals.
Compose The Right Resources at the Right Time for the Right Job
Broaden access to expensive resources like GPUs with the ability to dynamically reconfigure them on the fly based on the differing workload needs of students and researchers. The minute class is over, the exact same infrastructure can be reconfigured to meet the needs of research workloads.
Break The Scalability Limits of Server Hardware
Start small with an appliance consisting of four AMD-powered servers, a management node, two FabreX switches, and an Accelerator Pooling Appliance (APA) populated with up to 8 PCIe devices (GPUs, DPUs, FPGAs – whatever your workload requires, in any combination). Configure those resources into the compute instances required by different workloads without being bound by server sheet metal.Â
Scale the appliance to populate a full rack with up to 15 servers and 6 APAs, for a total of 48 PCIe devices that can be dynamically configured and reconfigured using NVIDIA Bright Cluster Manager. You now have a rack-scale server where the compute unit is the rack, and everything operates at the same full performance and latency as if the components were inside the server sheet metal.
Providing Turnkey, Multi-tenancy Workflows for Higher Education
Centralize, standardize, share, and optimize infrastructure across the varied workloads required by a university — both classroom and research settings.
Choose Your Configuration
Condo 1: Start Small
Choose your base configuration. The Condo 1 appliance delivers the power of 8-16 GPUs/FPGAs across 4-5 application hosts, and can accommodate over 100 students. These may be configured and reconfigured to align with the demands of the moment.
Condo 2: Scale Incrementally
Grow as you need by seamlessly moving from a Condo 1 to a Condo 2, which can deliver more than 200 students their own individual systems — including dedicated GPUs.
Condo 3: Teach & Compute at Scale
When your researchers need access to a massive bank of 48 GPUs for a single workload, upgrade to Condo 3. The same infrastructure will support over 300 students with their own individual systems, with the transition from workload to workload managed through NVIDIA Bright Cluster Manager.
What You Need When You Need It
It’s a promise! The GigaIO Composability Appliance, University Edition, Powered by AMD, leverages standard APIs and integration with industry-leading NVIDIA Bright Cluster Manager orchestration software to streamline how server instances are configured and reconfigured, without any forklift upgrades.
How it Works
The GigaIO Composability Appliance combines NVIDIA Bright Cluster Manager’s ability to easily build and manage clusters with GigaIO’s ability to connect AMD accelerators, servers, and other devices in a seamless dynamic fabric. 
Ease of use is at the forefront of this design, as its goal is to be used in a class/lab setting without requiring dedicated IT expertise.
- Easy-to-deploy, turnkey appliance for Higher Education
- Centralize, standardize, share, and optimize infrastructure
- Handle varied types of workloads across departments
Native integration with Bright Cluster Manager allows the owner to easily assign configurations prior to use, dividing hardware among students to allow them the experience of running actual simulation workloads on the same compute infrastructure they will utilize upon graduation.
Disaggregating resources outside of servers enables each department to share them. Each department can now compose systems precisely tuned to the needs of the applications they need to run.
Key Advantages
The GigaIO Composability Appliance: University Edition supplies students with access to equipment they will use in the real world, so they will be better prepared for the job market.
Flexibility and composability means that systems don’t remain idle while not being used for teaching — they can instead be reconfigured for actual simulation work and swapped back into teaching mode as needed. And of course the inherent benefit of using composable infrastructure is the ability to build heterogeneous configurations for varied types of workloads and trainings.
Other benefits include:
- Easy to deploy and maintain
- Short time to value
- Lower Total Cost of Ownership (TCO)
- Lower Research Cost of Ownership (RCO)
- Efficient
- Future-proofed
- Cloud-like agility
Resources
From Engineering.com: There is an Easy and Affordable Way to Improve HPC Performance
