Lisa Su has had a datacenter in her sights since taking office as president and chief executive officer of AMD in 2014, charting a course that will allow the processor to once again be a factor in the global server market after a decade or so. more than a few. following its brief but high profile success with its first Opteron processor in 2003.
Key to the journey are the Epyc X86 server processors, launched in 2017 with “Naples” and based on the vendor’s built-from-scratch Zen microarchitecture. AMD has delivered this in a rapidly changing enterprise compute market that includes the rise of hyperscaler cloud providers, data proliferation, greater focus on data movement and the emergence of artificial intelligence and machine learning.
The vendor also offers its Instinct GPUs for the datacenter and in February finally closed its long-awaited nearly $ 49 billion acquisition of Xilinx, bringing the company’s FPGA programmable logic and associated DSP engines, AI accelerators, memory controllers and the like. family of SmartNICs, designed to offload networking and storage functions to enable CPUs to better do their computing work.
Now AMD is about to take the final step to complete is the expanding family of datacenter offerings, announcing plans on Monday to buy Pensando and its programmable packet processor and distributed services platform for approximately $ 1.9 billion. The move will give the chip maker a platform used by cloud players like Microsoft Azure, IBM Cloud and Oracle Cloud, vendors like Hewlett Packard Enterprises ’Aruba Networks Business and larger businesses like Goldman Sachs.
Pensando is also working with VMware to integrate its platform with the virtualization vendor’s “Project Monterey,” an initiative announced in 2020 to deliver a consistent platform for distributed computing environments that run AI and machine learning workloads that use of data processing units (DPUs) to offload the hypervisor and other functions from CPUs to better SmartNICs that become the center of gravity between computing and networking. In particular, DPU computing engines can be used to offload networking (sometimes with literal packet processors) and storage workloads from the CPU while the software stack on the platform offers software- defined networking, storage and security services along with policy and management functions.
In Pensando’s case, its SoC includes 16 Arm cores and a P4 MPU that delivers Pensando services at 400Gb/sec. It is installed on any standard server with the company’s Distributed Services Card, providing server-side services and network visibility through its hardware bi-directional flow streaming and traffic monitoring. Policy and Services Manager delivers lifecycle management, security and visibility across the software stack, as well as integration with analytics, orchestration and management tools via APIs.
“At Pensando, we are adding another three columns with a unique packet-processing and distributed-services platform that allows us to speed up not only networking but security to speed up storage applications, to add new ability to monitor datacenter activity for threats. detection and performance optimization and do it all at once, ”said Forrest Norrod, senior vice president and general manager of AMD’s Datacenter Solutions Group, The Next Platform. “We think we’re completing the puzzle. We now have complete additions of CPU-, GPU-, FPGA-based acceleration and adaptable SoCs and finally, this piece that helps us combine them all in a secure, high-performance way that is in we think it will be. primarily to continue to allow our customers to measure, make it secure and do it efficiently by offloading a lot of infrastructure [tasks] which would otherwise consume CPU or other cycles. “
The deal is expected to close in the second quarter, after which Pensando co-founder and CEO Prem Jain and other companies will be included in Norrod’s group.
Pensando was founded in 2017 by a group of longtime Cisco Systems executives who made their names as part of a group of influential engineers who helped drive the networking giant’s “spin-in” strategy, where Cisco is the only investor in the startups they start and then buys the company at a hefty price once the product is built, folding it into Cisco’s portfolio. The group, which called MPLS as an acronym for their first names and founded companies such as Insieme Networks and Nuova Systems, also includes Soni Jiandani, co-founder and chief business officer at Pensando.
Jain, Jiandani, Mario Mazzola, and Luca Cafiero all left Cisco in 2016 after Chuck Robbins took over for longtime CEO John Chambers, who has since become CEO of JC2 Ventures and chairman of Pensando. The startup has raised $ 313 million since 2017, including $ 35 million in Series C funding last year. It has approximately 330 employees, with more than 200 in India. About two-thirds of Pensando’s workers are focused on software, according to Jiandani.
Jiandani told The Next Platform that from the beginning Pensando used the system in developing its product.
“This is how you put together very programmable [silicon] and do everything software-defined, “he said.” We made this system-on-a-chip, which is a real system that can run our software stack, it can run the logic of our customers and the two can live together with each other.It is form-factor agnostic, so if you put it on a server, we are eight to 13 times the performance of Amazon’s Nitro[the underlying platform for its EC2 instances]. The irony of it is that Amazon only uses it for running their own cloud and that’s why we say we’re doing cloud democracy. Whether you’re a high-end financial services organization like Goldman Sachs or you’re Microsoft Azure, we want to be able to deliver you eight to 13X performance steps and scale what a hyperscaler and lead in this space like AWS can afford. “
Last fall, Pensando used form factor-agnostic capabilities to introduce to Aruba the CX 10000, a top-of-rack networking switch that runs Aruba’s AOS-CX operating system and its Aruba Fabric Composer for CX provisioning switch. The switch is being shipped to thousands of Aruba customers, Jiandani said.
“That platform essentially offers the enterprise the same features that cloud customers like Microsoft and Oracle Cloud deploy to Pensando in size and we bring 100X the size to that smart switch, 10X the performance, and one-third the total cost of ownership, ”he said.
The programmability and SmartNIC capabilities of the Pensando platform have a similar chord to that offered by Xilinx. Norrod cited Xilinx SmartCards, saying they have been adopted by some of AMD’s larger enterprise customers. However, he says, the strategies taken by each company are different.
“Both solutions have a set of hard-coded accelerators for common features,” he said. “The difference is, for anything, at Pensando, we have a programmable, highly customizable P4 packet processor that can do many, many things and can be easily programmed into software to do many things. Xilinx has very high performance and, with Xilinx, you can do a lot of that and you can actually do it with higher performance, but you have to program the FPGA. The way we look at it here is that, for the latest performance for a narrow range of functions, we can address that using Xilinx technology and offer maximum performance for certain classes of workloads and certain specific customer. “
It also comes at a time when Intel-along with its infrastructure processing unit (IPU) and SmartNICs-and Nvidia (Bluefield DPU and DPU-based SmartNics) are both expanding their portfolios. None of them are the system software delivered by Pensando or the broader ecosystem, Jiandani says. Along with Aruba, Hewlett Packard Enterprise, and VMware and the cloud providers, the vendor is also partnering with Splunk, ServiceNow, Guardicore and Elasticity to help make Pensando a foundation in the zero-trust network architecture of enterprises and cloud providers.
“If I’m integrating within Oracle’s cloud environment and Microsoft’s cloud environment or I’m going into a Goldman-Sachs’ environment or if I’m in a smart-switch topology that fits Aruba’s customer environment, with ecosystem partners that this and a deep ability to have API-driven integration pointing to Splunk and others, we adapt to current models of operating customers in the enterprise and cloud, ”he said.
At Pensando, AMD will deliver a high-performance collection of silicon and software that can support a range of services simultaneously without having to do any programming such as the enterprise’s FPGA.
“It’s more accessible,” Norrod said. “You expect over time to combine the elements. I see them as very complementary now, but I also see them as very synergistic over time, so stay focused for a few years. It should be. you’d think we’d make some very interesting products that combined some of these things. “
“A customer can pick up the complete solution, they can replace their own piece or part of the networking function defined by the software if they have different needs,” he said. “They can mix and match and they can do it by running that code on the Arm core side of the park or on the programmable P4 packet engine. This is a very unique set of customization capabilities that Pensando offers to our customers. You’ll see a better roadmap over time. We’ve doubled and tripled networking and data processing with assets from Xilinx as well as Pensando. “