Sibley Highlights Power10’s Security and AI Capabilities
October 13, 2021 Alex Woodie
As the vice president and global offering management for Cognitive Systems at IBM, Steve Sibley often gets the honor of delivering the keynote address at the annual POWERUp conference. During his keynote at last week’s show, Sibley used the time to delve into the latest Power10 processor and its extensive list of new features, including some designed specifically to boost AI and security.
“One of the real privileges I have as project manager for Power Systems to be able to take the key innovations our team is developing in Rochester and our different development labs around the world, and bring it to you in new ways,” Sibley said during the opening session for POWERUp 2021, which COMMON delivered as a virtual event.
The “key innovation” that Sibley spent the bulk of his 35 minutes or so talking about was the Power10 processor, which IBM is currently shipping in the enterprise-class E1080 machine and will be rolling out across the entry-level and midrange machines over the next nine months.
The Power10 is an impressive feat of design and engineering, as all of IBM’s new Power processors tend to be. Sibley point out that, prior to the delivery of Power10, the top-performing platform in the world were those running Power9 processors. “So a lot of that is comparing against ourselves,” he said.
Of course, IBM competes with all enterprise-level chip and platform makers, most notably Intel, which still supplies the bulk of chips used to process transactions around the world. But it also competes with Nvidia, which develops the GPU processor at the heart of many machine learning training models, not to mention the OEMs like HPE, Dell, and Lenovo that build the servers that power transactional and analytic workloads, as well as the cloud hyperscalers (AWS, Microsoft Azure, Google Cloud) that are competing ferociously to absorb those workloads into their sprawling cloud data centers.
IBM has a captive audience with the IBM i platform, which Sibley quantified as being used by “tens of thousands of customers.” That captivity is sometimes good and sometimes bad. When the linkage is working well, it results in a tight feedback loop between the biggest IBM i customers and the top designers and decision-makers at IBM, who have shown themselves to be responsive to their customer needs. (When the linkage is not working well . . . well, that’s a story for a different day!)
As Sibley described the capabilities in Power10 and some of the other items that IBM has been working on, one thing that stood out was how the chip’s new design points are closely matched to address IBM’s customers’ concerns. Of course, everybody wants more power, and Power10 delivers that in droves (as all new generations of IBM’s processor do, in varying degrees). But Sibley’s presentation showed how IBM paid close attention to delivering some of the other capabilities that customers have asked for, such as bigger and better AI and security capabilities, which shows IBM does, indeed, listen to its clients.
Security has been a big focus area for IBM i shops and customers in general. Before the COVID-19 pandemic, security had been identified as the number one requirement for several years running, with system resiliency (i.e., high availability/disaster recovery) and application modernization typically taking the number two and three slots. If you look back through the archives, surveys by HelpSystems, Precisely, and IBM itself show this to be the case. (AI has not been a top concern, at least among IBM i shops.)
But something happened during COVID-19. The security threats were still there — indeed, we’ve seen the ransomware epidemic escalate rapidly. But according to Sibley, flexibility has overtaken security to be the number one concern.
“The emphasis around flexibility has really grown,” he said during the POWERUp keynote last week. “IBM does a business value study of thousands of CEOs and CIOs around the world and the thing that floated to the top this year is this whole area of operational agility and flexibility. In fact, 37 percent more respondents highlighted that as their top priority than even the next-most important priority, which happens to be security in their environment. So it’s within that backdrop that we want to be sure that we are continuing to deliver you the scalability, the flexibility, and agility on your platform and where and how you do business.”
IBM has several answers to the need for flexibility. Arguably the biggest offering in that department is IBM Cloud and the Power Systems Virtual Server, where customers can scale up and scale down IBM i, AIX, and Linux environments to their hearts content. Sibley also mentioned the new release of Access Client Solutions (ACS) and the new IBM i Navigator interface as delivering flexibility. But those don’t exist in the Power10 chip, and one can’t really say that the new capabilities they deliver were a direct response to the surging demand for flexibility. Red Hat Ansible modules and workbooks, similarly, provides the flexibility to manage IBM i environments without having IBM i-specific skills.
It will be interesting to see how IBM responds to its customers’ current demand for increased flexibility in the Power11 chip, which is currently under development. Composibility, or the ability for a platform to essentially shape-shift to adapt to changing workloads, is all the rage in the open systems world. Maybe it will come to IBM i too. (The PowerAXON interconnect and OMNI memory components on the outside of the Power10 chip certainly could lend themselves to greater flexibility in the I/O and compute/memory acceleration department.)
Clearly, as the Power10 chip was being developed, IBM heard its customers’ demands for greater security, which had been the number one customer concern for the last few years until COVID-19 dislodged it from the top spot and made flexibility the top concern (at least according to one IBM survey). It delivered several compelling new capabilities in that regard.
“One of the key focuses that we had with Power10 was around security,” Sibley said. “In fact, we started with this focus of ensuring best overall platform protection, because security does take on many layers.”
Part of that includes encrypting the content of memory on the Power10 chip, which can prevent hackers from reading data residing in IBM i, AIX, and Linux partitions.
“So we built in what we call transparent memory encryption,” Sibley said. “You don’t have to do anything special from a workload standpoint other than just turn it on. It has no performance impact to be able to encrypt all their memory on your platform.”
IBM also added four new cryptography engines with Power10 to provide encryption for data in motion, and took steps to defend against return-oriented programming attacks on the processor, which were personified by the Spectre and Meltdown vulnerabilities that impacted Power and Intel chips in early 2018. IBM also developed isolation capabilities for the service processors, and implemented new “trusted boot” capabilities, Sibley said.
“But it’s not just about the platform protection,” he continued. “It’s really about an integrated approach to security.” To that end, he mentioned the real-time intrusion connection in Power SC product, as well as built-in multi-factor authentication. Sibley also mentioned new IBM i Services IBM in the latest Technology Refreshes for IBM i 7.3 and 7.4 that delivers a dashboard view of current security status across IBM i platforms, and the IBM Safeguarded Copy capability in Flash Systems, which protect against ransomware.”
“We encourage you to take this comprehensive view of security in the platform,” he said. “We’re committed to giving you the most secure platform overall, the tools to help you from a monitor and detection standpoint.”
The Power10 processor also ups the ante in the AI department, which is more of an offensive capability designed to make businesses run better (although it can also be used to fight security threats and fraud, which gives it a defensive angle, too). If Nvidia’s soaring stock price over the past five years is any indication, AI is a growing workload, and it’s one that IBM aims to capture more of with its Power platform.
While IBM i shops aren’t likely to be training massive deep learning models on IBM i LPARs, the demand for better AI capabilities comes through in other ways, according to Sibley. To that end, IBM bolstered the performance of matrix math acceleration on Power10, which can be used with transactional data for workloads such as supply chain planning and pricing analysis, Sibley said.
“We are enabling all of the key runtime libraries right on the platform itself, either on Linux, where they are naturally developed initially, but also working within IBM i and AIX to bring it right on the partition itself, to support those libraries and AI frameworks from a runtime standpoint,” Sibley said. “That means you can train your AI models anywhere, from the cloud or a different platform, but then bring the execution, the inferencing capabilities of that, right onto your core platform, your core application and data itself, to get the performance you need.”
The built-in AI acceleration on the Power10 chip will pay dividends in other ways. According to Sibley, it will lessen the need to move data and workloads to other platforms. “With Power10, they’ll actually be able to [run AI workloads] more integrated into fewer systems because of the acceleration that we built into the Power10 processor itself, as opposed to offloading some of this from one Power System to another Power System,” he said.
Minimizing the movement of data also brings the side benefit of reducing the security exposure, which is another plus of this approach.