Archive for the 'Cloud' Category
-
June 18, 2024
Custom Compute in the AI Era
This article is the final installment in a series of talks delivered Accelerated Infrastructure for the AI Era, a one-day symposium held by Marvell in April 2024.
AI demands are pushing the limits of semiconductor technology, and hyperscale operators are at the forefront of adoption—they develop and deploy leading-edge technology that increases compute capacity. These large operators seek to optimize performance while simultaneously lowering total cost of ownership (TCO). With billions of dollars on the line, many have turned to custom silicon to meet their TCO and compute performance objectives.
But building a custom compute solution is no small matter. Doing so requires a large IP portfolio, significant R&D scale and decades of experience to create the mix of ingredients that make up custom AI silicon. Today, Marvell is partnering with hyperscale operators to deliver custom compute silicon that’s enabling their AI growth trajectories.Why are hyperscale operators turning to custom compute?
Hyperscale operators have always been focused on maximizing both performance and efficiency, but new demands from AI applications have amplified the pressure. According to Raghib Hussain, president of products and technologies at Marvell, “Every hyperscaler is focused on optimizing every aspect of their platform because the order of magnitude of impact is much, much higher than before. They are not only achieving the highest performance, but also saving billions of dollars.”
With multiple business models in the cloud, including internal apps, infrastructure-as-a-service (IaaS), and software-as-a-service (SaaS)—the latter of which is the fastest-growing market thanks to generative AI—hyperscale operators are constantly seeking ways to improve their total cost of ownership. Custom compute allows them to do just that. Operators are first adopting custom compute platforms for their mass-scale internal applications, such as search and their own SaaS applications. Next up for greater custom adoption will be third-party SaaS and IaaS, where the operator offers their own custom compute as an alternative to merchant options.
Progression of custom silicon adoption in hyperscale data centers.
-
June 11, 2024
How AI Will Drive Cloud Switch Innovation
This article is part five in a series on talks delivered at Accelerated Infrastructure for the AI Era, a one-day symposium held by Marvell in April 2024.
AI has fundamentally changed the network switching landscape. AI requirements are driving foundational shifts in the industry roadmap, expanding the use cases for cloud switching semiconductors and creating opportunities to redefine the terrain.
Here’s how AI will drive cloud switching innovation.
A changing network requires a change in scale
In a modern cloud data center, the compute servers are connected to themselves and the internet through a network of high-bandwidth switches. The approach is like that of the internet itself, allowing operators to build a network of any size while mixing and matching products from various vendors to create a network architecture specific to their needs.
Such a high-bandwidth switching network is critical for AI applications, and a higher-performing network can lead to a more profitable deployment.
However, expanding and extending the general-purpose cloud network to AI isn’t quite as simple as just adding more building blocks. In the world of general-purpose computing, a single workload or more can fit on a single server CPU. In contrast, AI’s large datasets don’t fit on a single processor, whether it’s a CPU, GPU or other accelerated compute device (XPU), making it necessary to distribute the workload across multiple processors. These accelerated processors must function as a single computing element.
AI requires accelerated infrastructure to split workloads across many processors.
-
March 23, 2023
How Secure is Your 5G Network?
By Bill Hagerstrand, Security Solutions BU, Marvell
New Challenges and Solutions in an Open, Disaggregated Cloud-Native World
Time to grab a cup of coffee, as I describe how the transition towards open, disaggregated, and virtualized networks – also known as cloud-native 5G – has created new challenges in an already-heightened 4G-5G security environment.
5G networks move, process and store an ever-increasing amount of sensitive data as a result of faster connection speeds, mission-critical nature of new enterprise, industrial and edge computing/AI applications, and the proliferation of 5G-connected IoT devices and data centers. At the same time, evolving architectures are creating new security threat vectors. The opening of the 5G network edge is driven by O-RAN standards, which disaggregates the radio units (RU), front-haul, mid-haul, and distributed units (DU). Virtualization of the 5G network further disaggregates hardware and software and introduces commodity servers with open-source software running in virtual machines (VM’s) or containers from the DU to the core network.
As a result, these factors have necessitated improvements in 5G security standards that include additional protocols and new security features. But these measures alone, are not enough to secure the 5G network in the cloud-native and quantum computing era. This blog details the growing need for cloud-optimized HSMs (Hardware Security Modules) and their many critical 5G use cases from the device to the core network.
-
January 04, 2023
Software-Defined Networking for the Software-Defined Vehicle
By Amir Bar-Niv, VP of Marketing, Automotive Business Unit, Marvell and John Heinlein, Chief Marketing Officer, Sonatus and Simon Edelhaus, VP SW, Automotive Business Unit, Marvell
The software-defined vehicle (SDV) is one of the newest and most interesting megatrends in the automotive industry. As we discussed in a previous blog, the reason that this new architectural—and business—model will be successful is the advantages it offers to all stakeholders:
- The OEMs (car manufacturers) will gain new revenue streams from aftermarket services and new applications;
- The car owners will easily upgrade their vehicle features and functions; and
- The mobile operators will profit from increased vehicle data consumption driven by new applications.
What is a software-defined vehicle? While there is no official definition, the term reflects the change in the way software is being used in vehicle design to enable flexibility and extensibility. To better understand the software-defined vehicle, it helps to first examine the current approach.
Today’s embedded control units (ECUs) that manage car functions do include software, however, the software in each ECU is often incompatible with and isolated from other modules. When updates are required, the vehicle owner must visit the dealer service center, which inconveniences the owner and is costly for the manufacturer.
-
November 28, 2022
A Marvell-ous Hack Indeed – Winning the Hearts of SONiC Users
By Kishore Atreya, Director of Product Management, Marvell
Recently the Linux Foundation hosted its annual ONE Summit for open networking, edge projects and solutions. For the first time, this year’s event included a “mini-summit” for SONiC, an open source networking operating system targeted for data center applications that’s been widely adopted by cloud customers. A variety of industry members gave presentations, including Marvell’s very own Vijay Vyas Mohan, who presented on the topic of Extensible Platform Serdes Libraries. In addition, the SONiC mini-summit included a hackathon to motivate users and developers to innovate new ways to solve customer problems.
So, what could we hack?
At Marvell, we believe that SONiC has utility not only for the data center, but to enable solutions that span from edge to cloud. Because it’s a data center NOS, SONiC is not optimized for edge use cases. It requires an expensive bill of materials to run, including a powerful CPU, a minimum of 8 to 16GB DDR, and an SSD. In the data center environment, these HW resources contribute less to the BOM cost than do the optics and switch ASIC. However, for edge use cases with 1G to 10G interfaces, the cost of the processor complex, primarily driven by the NOS, can be a much more significant contributor to overall system cost. For edge disaggregation with SONiC to be viable, the hardware cost needs to be comparable to that of a typical OEM-based solution. Today, that’s not possible.
Recent Posts
- 2006: The Twelve Months That Changed the Chip Industry
- Encrypting Data in Use: How HSMs Can Advance Next-Generation Confidential Compute
- Nine Things to Remember About the Future of Copper in Computing
- Cloud-Managed Enterprise (CME) Switches Powered by SONiC
- Marvell COLORZ 800 Named Most Innovative Product at ECOC 2024
