Intel makes zillions of processors that are used worldwide. Still, it is also filling a need in another burgeoning market with a smart-PCI Express (peripheral component interconnect express) card: Intel® NetSec Accelerator Reference Design for network security providers. The reference design, last updated in November, contains an Intel® Ethernet Controller E810 combined with an Intel® Xeon® D processor, configured in a PCIe add-in card.
These units serve as autonomous and independent servers at the edge. They are located in the space you find an Ethernet network interface card and enable security developers to speed up integration and delivery of high-quality network security solutions at the edge with the versatility to handle all required feature sets. The accelerator card comes in both 8- and 16-core CPU SKUs, meaning a dual card deployment can scale the compute of an existing host by up to 32 cores.
In short, we’re talking about a solution that accelerates speeds in network infrastructure, freeing up CPU cores for better and more efficient application performance. These accelerator add-in-cards can be installed into PCIe slots of Intel servers or network and network security appliances built on Intel platforms. Each card acts as a single, autonomous server and is easily swappable for specific use cases.
These PCIe cards augment the compute and I/O to meet the evolving network security requirements in the modern edge to cloud workload environment.
Supermicro comes to the fore with an agile approach
Supermicro is a respected, well-established IT hardware manufacturer, having been in business since the dawn of the internet in 1993. The San Jose, CA-based company has always been valued as the type of independent server OEM that plays well and connects smoothly with other parts and application makers. Supermicro’s innovative SLED-based (slot-loadable enclosable device) server architecture enables compute and security acceleration through its modularity.
Each SLED can hold different modules, such as compute modules or accelerator cards. This allows users to start with only the resources they need and then scale up or down by adding more SLEDs or swapping SLED components.
SLEDs can hold multiple NetSec PCIe accelerator cards, based on the aforementioned Intel NetSec Accelerator Reference Design, which provide high-level security and networking processing capabilities. The accelerator card offloads work from CPUs and accelerates tasks such as AI inferencing for network security, which is rapidly becoming a standard use case in many security solutions for vertical industries. Thus, the SLED-based design allows flexible deployment of acceleration resources close to the workload – and the closer the workload to its data, the more efficient the workflow will be.
Supermicro has partnered with Intel for many years and offers a range of server models that utilize Intel reference designs and technologies. These solutions cater to various workloads and applications, including cloud computing, AI/ML, storage and edge computing.
“With Intel NetSec Accelerator Reference Design, you’re adding more functionality, although it’s essentially still a NIC,” says Jeff Sharpe, Director of AI Solutions at Supermicro. “The host OS will always see it as that, except now you have extra processing power that’s tuned and reference-architected to deliver an optimized experience, whether it’s AI, whether it’s networking, security and other capabilities.”
Sab Gosal, Sr. Director of Intel’s Network and Edge Group, said that Supermicro is ideal to work with because “it has all SLED-based architectures. So within one type of footprint, you can have three to four SLED compute sites or service sites that you can plug in. Modularly goes up and down. Within those slides, you can add accelerator cards or NICs. This one foundational architecture can be scaled in whichever way the consuming customer wants to deploy, based on their own needs. It’s beautiful.”
Sharpe said Supermicro sees “avenues of growth for enterprise customers and different levels from small, medium and all the way to telcos and large enterprises. The systems are all able to grow based on the size of workloads within each customer. The market for enterprise is growing, where you have enterprise integrators and telco systems supplying more and more enterprise-class products (developed on their own). This is where the true value of Intel NetSec Accelerator Reference Design comes into play.”
One advantage a smart NIC enables is a front-end processor that manages packets, both incoming and outgoing. That enables a network administrator to put additional software assets on the NIC card that offloads the workflow from the CPU, which can — and should — be doing other heavier functions.
“This is especially the case as we’re getting into AI, model delivery and AI inferencing functions, where you want your CPU to be working more of its functionality of those key app attributes versus doing networking and security that you want your NIC processor to do,” Sharpe said.
The true value of the combination of Intel and Supermicro is about hardware options that can be deployed at the edge, from a small system to an extra-large system, such as AWS or Google, Sharpe said. “With the latest technology this provides, we can do additional security and networking functions on that system while offloading the core processor to do its own thing for higher-end functions,” he said.
Versatility in all its forms and apps
Because this technology now is interpluggable, interchangeable and repurposable, a network administrator can easily deploy any OS package he or she wants, so this creates an opportunity for companies such as Supermicro or Versa to triangulate, Gosal of Intel said.
“That’s just one use-case example,” Gosal said. “Let’s now think of a modular server where one SLED could be one customer that you’re hosting, and it could be a SASE POP (point of presence). The next one could be a 5G POP, and so forth. Another could be a NGFW (Next Generation Firewall), all hosted in one environment.”
As the industry moves forward, OPEX metrics (involving colocation facilities) are beginning to become a major business factor, Gosal said.
“As you know, infrastructure is growing. Users need to get more out of their footprint and add a better sort of power efficiency – a sustainability type of metric, that is a lot more ergonomically and environmentally aligned. We’re really excited about this lineup because it is the fundamental recipe for edge networks,” Gosal said.
Sharpe laid out the high level of this new market from Supermicro’s perspective.
“Our marketplace is mostly about small and medium-sized businesses, but we do have large enterprise companies as well doing edge-to-cloud technologies,” Sharpe said. “It’s retail, smart cities, smart places. And I think we’ve turned these Points of Presence where the servers go from individual systems in small businesses to micro-clouds or data centers at the edge. But this is primarily focused at the edge and using edge-to-cloud infrastructure for those applications.”
Intel and Supermicro use case documentation
Supermicro offers server packages that utilize all of Intel’s newest reference designs and technologies. They include:
- Supermicro X13 servers. The recent release of the latest Supermicro X13 server product lines with 5th Gen Intel® Xeon® Scalable processors enables enterprises and cloud providers to upgrade their existing infrastructure to new servers and quickly realize performance gains. Everything is hot-swappable for quick changes.
- Supermicro Cloud Media and AI Servers with Intel GPUs. Optimized for media and cloud workloads, Intel GPUs comprise an open, standards-based software stack. These solutions also incorporate essential server capabilities, ensuring high reliability, availability and scalability across various applications, including media processing, delivery, AI visual inference, cloud gaming and virtualization.
- Supermicro ‘Accelerate Everything’ roadmap. Supermicro has long designed and manufactured a range of servers that address a range of workloads. Intel’s general-purpose CPUs (the latest (4th and 5th Gen Intel Xeon Scalable processors) fit well with Supermicro’s Accelerate Everything approach. Specially designed silicon now accelerates specific workloads, resulting in multiples of performance for these applications.
This post was originally published on the 3rd party site mentioned in the title of this this site