Power Systems Roadmap Focuses on AI, Cloud and Security
Disruptors to business are everywhere. Startups challenge the status quo. Artificial intelligence changes how an organization gains insight. Solving these challenges requires innovation.
Image by Script & Seal
By Shirley S. Savage12/01/2018
Disruptors to business are everywhere. Startups challenge the status quo. Artificial intelligence (AI) changes how an organization gains insight. Hackers constantly probe enterprise IT to detect and exploit security gaps. Solving these challenges requires innovation, cutting-edge technologies and, in many cases, fundamentally rethinking the current approach.
IBM firmly believes an exponential shift is occurring in the way business operates. That shift is centered on a data-centric approach, says IBM Chairman, President and CEO Ginni Rometty. Enterprises have a wealth of data within their businesses. Eighty percent of the world’s data resides inside enterprise IT and isn’t searchable on the web, she says. Strong security defenses and data protections are essential to exploit this treasure trove of data.
Organizations are using AI for exponential learning. “Businesses are embracing their new capability to learn in every part of their operations, from the back office to the front office, for a simple reason: They understand that the ultimate competitive advantage is to out-learn your competitors,” notes Rometty in a keynote address at Think 2018.
As companies garner the benefits of exponential learning based on their own data, it will alter who becomes a disruptor. Disruption will come from incumbent companies, not new entrants. These “incumbent disruptors” will leverage intelligent digital platforms to unlock insights, Rometty points out. Those insights will change the way the world works.
Incumbent disruptors will use cloud platforms to create new services and revenue streams. Private, public or hybrid cloud will be essential to how business is done.
“In a world where data is the basis for competitive advantage, businesses need a partner they trust,” Rometty says. IBM and IBM Cognitive Systems are committed to helping clients embrace change and thrive with innovations in AI, cloud and security.
IBM created a Power Systems* Roadmap that puts clients running AIX*, IBM i and Linux* on POWER* in the forefront of these exciting innovations.
AI has gone from a sci-fi concept to reality—made possible by accelerated computing. The Summit supercomputer at Oak Ridge National Laboratory utilizes the IBM Power Accelerated Computing Platform to provide unprecedented performance for AI, analytics and high-performance computing.
AI and related workloads—machine learning and deep learning—require accelerators to function optimally. IBM has been working with accelerators for some time and clients are reaping the benefits. “IBM has added to the highly differentiated accelerated ecosystem in the form of NVLink 2.0 to connect to GPU and OpenCAPI*, delivering 10x CPU to accelerator bandwidth,” says Satya Sharma, CTO, IBM Power Systems.
IBM is developing the future generations of NVLink and OpenCAPI interfaces to deliver more bandwidth and expand system coherence for ease of programming. “Those attributes will continue to offer sustained differentiation versus x86,” he says. “We believe Power Systems is the right architecture to take clients into the next era of IT innovation.”
The classic enterprise and AI, machine learning and deep learning worlds are moving closer together. Right now, AI is well established in imaging, video, natural language and speech recognition, and related categories. “Over the coming years, AI will expand to mainstream enterprise applications across all of the major industries,” Sharma comments. “As enterprise applications become AI aware, deep integration will be needed for training space with enterprise structured data. So much of that data resides on Power Systems.”
Enterprise applications will incorporate AI inferencing as transactions are processed. The POWER10 processor will implement additional AI capability to enable use of in-line AI integration. “IBM Cognitive Systems will be at the forefront of the innovation as these worlds come together,” Sharma contends.
"Over the coming years, AI will expand to mainstream enterprise applications across all of the major industries."–Satya Sharma, CTO, IBM Power Systems
As businesses seek to exploit AI, machine learning and deep learning workloads, the IBM POWER9* processor’s accelerated systems reduce training time to improve data science productivity. IBM’s PowerAI software fully supports open-source frameworks such as TensorFlow and Caffe, making it easy for clients to launch AI applications.
A data-centric focus for business is reliant on good security for data and operations. IBM has always approached security as a top concern and area of innovation. This view is now more pertinent than ever. CIOs and CISOs realize that new workloads provide great benefits for the organization but also add security risks, says Dimitrios Pendarakis, chief security officer, IBM Cognitive Systems.
In the past, a tight perimeter around enterprise IT assets, usually the servers, clients and storage systems was considered sufficient. “With the emergence of clouds services, mobile platforms and bring your own device (BYOD) policies, the boundaries between the personal and the enterprise are blurred,” Pendarakis notes. The potential avenues to security attacks increase as an employee can connect a personal iPad to the network to access internal documents and collaboration applications.
Prevention is necessary because breaches are costly. The “2018 Cost of Data Breach Study: Global Overview,” published by IBM Security and Ponemon Institute, shows that the average total cost of a data breach is $3.86 million, a rise of 6.4 percent from 2017 (ibm.com/security/data-breach). The average cost per lost or stolen record increased 4.8 percent to $148 in 2018. The study found a 27.9 percent likelihood of a recurring material breach over the two years, up from 27.7 percent in 2017.
Further, security of hardware and firmware needs to be addressed. Organizations are realizing that security needs to apply across the entire stack not just software. “System vulnerabilities impact fundamental pieces of computing and must be guarded against,” Pendarakis says.
New workloads such as AI create additional security challenges, requiring IT to secure inferencing systems, training and data results. “IBM Cognitive Systems is working with the IBM Security business unit, IBM Research, IBM Watson*, IBM Cloud* and other partners to see how system capabilities can help provide integrity and confidentiality of both AI data and models,” Pendarakis comments.
“We are also looking at issues like accountability of AI system decisions, eliminating potential bias in AI models and isolation, to make sure these AI workloads aren’t affected by other applications that are running on the system,” he says.
Every security tool helps protect data. However, the one wild card in the stack is the human factor. “While there can be malicious or intentional misbehavior on the part of employees, one of the biggest security concerns is still inadvertent errors by insiders,” states Pendarakis. IBM works to ensure security takes into account rogue actions as well as human mistakes.
IBM is working on new security features that cover both the compute platform, as well as workloads such as LPARs and virtual machines (VMs):
Secure Boot is designed to protect system integrity by verifying all firmware components via digital signatures.
Trusted Boot ensures that only authorized firmware components are being run on a server.
IBM also is enhancing built-in cryptographic offerings to include continuous encryption of sensitive application workloads and data.
The Power of Partnerships
IBM Power Systems clients benefit from an open ecosystem with partners. With security in mind, IBM sees to it that all elements work together. Security is infused across the entire stack from chip design, testing, interfacing with peripheral components, memory, firmware and OSes.
“At IBM, everyone is security cognizant,” states Pendarakis. “There are security experts across every component and design that IBM makes.”
IBM is dedicated to open source and works closely with the Linux community on security. POWER9 hardware and firmware are constantly being improved for secure cloud deployment with key features for both PowerVM* virtualization and OpenPOWER servers. IBM is also introducing a hardware isolation mechanism called Protected Execution Facility (PEF) for OpenPOWER servers. PEF provides additional security for VMs. It will be based on a combination of small processor extensions and open source firmware, which means clients and partners can test, verify and contribute to it.
"IBM provides end-to-end security from the processor to the OS and everything in between, either directly or via partnerships with the OpenPOWER ecosystem."–Dimitrios Pendarakis, chief security officer, IBM Cognitive Systems
For instance, with competitor’s systems, different vendors supply the hardware, processor, firmware and OS. But that’s not the case when a client chooses IBM Power Systems. “IBM provides end-to-end security from the processor to the OS and everything in between, either directly or via partnerships with the OpenPOWER ecosystem,” says Pendarakis. “This is unique to IBM.”
Now and in the future, Power Systems clients can expect openness and transparency with systems and security. Add-ons like IBM PowerSC* security and compliance, as well as PowerSC Multi-Factor Authentication, specifically manage security features. IBM also is using some of the system security technologies to secure and harden AI workloads. “IBM is willing to address any security problems that clients have,” Pendarakis says. “We’re here to understand the risks and provide solutions,” he notes.
IBM’s roadmap for Power Systems includes new cloud offerings. For on-premise clouds, IBM currently offers IBM PowerVC* and IBM Cloud Private Solutions. In addition, IBM is partnering with VMWare to enable vRealize on the Power Systems platform, says Sharma.
For the public cloud, IBM will be unveiling IBM Cloud on POWER for Linux, AIX and IBM i. Clients will be able to get the same security features in the cloud as available on-premise, notes Pendarakis.
Power Systems servers provide a highly secured server platform. Substantial improvements are being made to POWER9 hardware and firmware to provide enhanced security for cloud deployment with key features for PowerVM and OpenPOWER servers.
IBM is committed to Power Systems and to ensure the platform remains innovative. “IBM will continue to bring new technologies and form factors to market that address client needs,” says Sharma.
System capacity growth will benefit AIX, IBM i and Linux workloads for clients. For example, SAP HANA is a very resource-demanding workload particularly in terms of memory capacity. “We’re focused on a number of advanced memory techniques including hybrid memory to slow down the memory costs and vastly improve economics,” he notes.
IBM’s investment includes all three OSes: AIX, IBM i and Linux. It’s interesting to note that Linux on POWER now represents about 25 percent of Power Systems revenue. “That’s not because AIX and IBM i business has declined,” Sharma points out. “All three OSes are performing well.”
Moreover, the POWER9 processor provides a 50 percent increase in capacity versus IBM POWER8* for a broad spectrum of workloads. It uses NVLink 2.0 to connect GPA and OpenCAPI and delivers 10x CPU to accelerator bandwidth. POWER9 is the first processor to utilize the PCI Gen 4 I/O bus, doubling I/O bandwidth. “Many of these innovations have come about as a result of OpenPOWER industry collaboration with Google, NVIDIA, Mellanox and more than 300 other members,” Sharma says.
IBM and its ecosystem partners continue to develop and bring to market innovations that change the way organizations do business, stay secure and incorporate new workloads.
Shirley S. Savage is a writer and communications strategist. She's fascinated by tech, science, finance, energy and the way innovative people think.
Post a Comment
Note: Comments are moderated and will not appear until approvedcomments powered by Disqus