POWER > Business Strategy > POWER8

Competitive Intelligence

Kim Storin and Adel El-Hallak
Vice President, Marketing, IBM Cognitive Systems Adel El-Hallak, Director, Product Marketing, IBM Cognitive Systems-Photo by Matt Carr

“The key to success is the IT department stepping up and becoming a deep learning hero,” says Adel El-Hallak, director, Product Marketing, IBM Cognitive Systems. “It’s a shift from the cost-effectiveness mindset to, ‘How do we enable new revenue streams or business models that equip us to do things we couldn’t do before?’ ”

The cognitive transformation starts with understanding the distinctions among deep learning, machine learning, AI and advanced analytics. It expands with determining when to tap into on-premises data versus public cloud data. “Understanding these capabilities enables IT to have a seat at the table and to really help drive competitive advantage for their company,” El-Hallak adds. “The world is moving to AI and IT needs to be an integral part of that effort.”

AI can also be applied to IT challenges. For example, deep learning techniques can enable a cloud-based network to monitor cluster health and predict when a server might be in danger of going down. The system can be taught to recognize key signs of trouble and fix those issues before the asset fails. It can meet demand spikes in real time and mine structured and unstructured data, this time to address IT questions for critical applications.

On-Premises Solutions

Cognitive computing is frequently associated with the public cloud, or delivered as a service in the form of applications such as IBM Watson*. Cloud deployments reduce the amount of resources and in-house expertise required to execute this technology, making them a suitable entry point for many organizations. At the same time, public-cloud solutions aren’t appropriate for proprietary data that’s too sensitive to transmit. On-site deployment in a private cloud is an increasingly accessible option, especially as deep learning applications move from development to actual production.

“We talk about cognitive computing as a public cloud application so often that it’s easy to forget that enterprises can build and deploy these capabilities in-house,” Storin says. “There’s so much data that lives within the organization that we don’t exploit to its full capability.”

Ready for the Cognitive Era

Enter the IBM Power Systems* platform, designed from the silicon on up to handle compute-intensive workloads and massive amounts of data. The POWER8* chip boasts twice the number of cores as x86 servers, along with as much as five times the memory bandwidth and three times the memory cache. It leverages high-speed I/O, hardware accelerators designed to handle specific workloads, Coherent Accelerator Processor Interface (CAPI) technology to provide a pipeline directly to the chip and more. Collectively known as POWERAccel, these functionalities are designed to meet the cognitive computing challenge.

In order to leverage Power Systems’ best-in-breed accelerator technology, IBM helped found the OpenPOWER Foundation, an open technical community that enables collaborative development on top of the Power Systems architecture to solve the toughest problems of our digital age.

To help enterprises tap into the power of deep learning, IBM created PowerAI, a software distribution that’s easy to deploy across a private cloud. By combining this software with IBM Power Systems hardware, enterprises can rapidly deploy a fully optimized and supported deep-learning platform on premises. PowerAI includes major open-source deep-learning frameworks such as Tensorflow and Caffe. It’s optimized to run on the Power Systems accelerated system for AI, which boasts a CPU-to-GPU NVIDIA NVLink interface. The Linux* software binary can be deployed across a single system or across a cluster or scale-out model. Businesses can also take advantage of the Nimbix Cloud, which hosts Power Systems servers and PowerAI.

Cognitive computing provides tools to enable organizations to serve clients and compete in the marketplace in new ways. IBM supports the effort by providing a choice of implementations. Public cloud or private cloud, operated by the IT shop or hosted in the public cloud by an IBM partner, this suite of cognitive computing options gives modern enterprises the tools they need to lead the market.

Leveraging AI

The total amount of data in the world is expected to reach 44 ZB by 2020. To gain benefit, organizations must do more than just manage the flood of bits and bytes. They need an infrastructure and an AI strategy that will enable them to serve clients and gather business insights on a near-real-time basis.

And with its ability to tackle new problems in innovative ways, AI promises to remake the competitive landscape. “There are a lot of people out there like data scientists, developers and IT leaders who are exploring machine learning and different types of use cases,” Storin says. “There’s a sense that we haven’t even scratched the surface of the capabilities that AI can bring to the table.”

Kristin Lewotsky is a freelance technology writer based in Amherst, N.H.

comments powered by Disqus



2017 Solutions Edition

A Comprehensive Online Buyer's Guide to Solutions, Services and Education.


Another Lifeline for those on AIX 5.3 Extended Support


Competitive Intelligence


Coverage of IBM's OpenPOWERSummit Announcements

IBM Systems Magazine Subscribe Box Read Now Link Subscribe Now Link iPad App Google Play Store