Part of the Technology photoes in this website are created by rawpixel.com - www.freepik.com

Composable solution for Edge AI inference- Reducing Infrastructure Footprints as Services Expand

992

Artificial intelligence (AI) is growing rapidly in the past few years and is applied to all kinds of daily life services. Today, we have autonomous driving cars, interactive voice response systems, precise weather forecasts, and advanced financial and medical services backed by AI. Nonetheless, the key to provide reliable services and instant feedback is the computing systems that perform real-time inference behind the scenes.

Following the maturity of 5G network, the demand for AI inference at the edge will rapidly increase for more reliable and timely services, in which drives the need for decentralized computing facilities to handle large amount of data. The larger channel and higher speed certainly benefit the service providers; however, one big challenge emerges: How to minimize the infrastructural cost as the service coverage area expands?

To respond the aforementioned challenge, H3 Platform introduces the Falcon 41 series PCIe Gen 4 composable disaggregated solution. Falcon 4109 (9-slot model) and 4118 (18-slot model) are specially designed for edge computing applications and micro data center. The Falcon 41 series possesses the great flexibility and scalability that any edge computing and micro datacenter needs as this composable solution can further increase the device density. “Enterprises will need strong computing power to speed up their services… stacking up servers as business grow wouldn’t work for long term, enterprises need a more efficient solution.”—said Brian, CEO of H3 Platform.

Being a composable disaggregated solution, the Falcon 41 series not only supports resource dynamic allocation among multiple hosts, but also provides more PCIe slots than any existing server equipment does. By disaggregating accelerator and storage devices from servers, the requirement of server itself will be significantly lower, thus narrowing down the use of bulky server chassis. Falcon 41 series helps IT personnel to increase the utility of devices and plan for more efficient deployments in limited space.

H3 Platform research and development department also noted that the company is aiming to incorporate and optimize DMA and SR-IOV technologies to its composable solution products. “The hardware aspect of our solution is pretty matured…and these new features will be accessible with our future software updates.”

As AI as a service rises, more and more IT facilities are going to be built to support edge AI inference. It is critical for enterprises to reconsider the economic and environmental impacts of the increasing IT footprints.


category : GPU
tags :