Cisco is expanding its Unified Computing System (UCS) to help organizations take full advantage of the growing demand for machine learning and artificial intelligence applications.
The new Cisco UCS C480ML server platform is Cisco's first system purpose-built for machine learning. It's a system that benefits from both hardware and software needed to accelerate machine learning workloads.
On the hardware side, the UCS C480 ML Rack Server is a 4 Rack Unit (RU) form factor server that differs from the existing UCS C480 M5 server in a number of critical areas.
The ML edition supports up to eight NVIDIA Tesla v100 GPUs that are all connected with the NVIDIA NVLink Interconnect.
In terms of CPU power, the C480 ML is powered by a pair of Intel Xeon Scalable processors, with up to 28 cores per socket. The 4RU chassis can also be configured with up to 3 TB of memory, spread across 24 DDR4 DIMMs. Storage on the UCS 480 ML is also large, with support for up to 24 SAS/SATA-based SSD/HDDs and up to 6 NVMe drives.
Beyond just having a beefy hardware server platform, Cisco has partnered with leading vendors and efforts in the space to create a series of Cisco Validated Designs. Among the partners is Cloudera and its Data Science WorkBench platform. Cisco is also partnering with Cloudera's rival Hortonwork for a validated design on the Hortonworks Hadoop 3.1 Data Lake.
Additionally, Cisco has been working with Google to help enable the open-source Kubeflow project for the C480 ML. Kubeflow is an attempt to make it easy for organizations to run the Tensorflow machine learning framework inside of a Kubernetes container orchestration system.
"The impetus for us to develop the system has really been pushed from our UCS install base, customers, including many in finance, public sector and healthcare," Todd Brannon, Sr. Director of Data Center Marketing at Cisco, told ServerWatch.
Brannon added that organizations are increasingly moving past just doing analytics into looking at machine learning and deep learning, which unlocks the potential for machines to perform complex tasks at large scale.
"For us entering this market, it's about bringing the acceleration technology that they need in a way that's very non-disruptive," Brannon said.
Sean Michael Kerner is a senior editor at ServerWatch and InternetNews.com. Follow him on Twitter @TechJournalist.