China Mobile recently gathered the leading 5G hardware and software vendors to an “Open User Plane Function” conference, where they addressed the need to create a universal platform for user plane functionality (UPF), which is the data plane within a 5G network.

In other words, in the opinion of one of the world’s largest telecommunications operators, UPF appliances need to be both open and decoupled from the entire 5G infrastructure, in particular the SMF (session management function) block, which sends control information to the UPF. Keeping the control and data planes separate increases the flexibility of the network infrastructure and facilitates operation at the network edge.

China Mobile took this initiative because it recognizes that UPF can make or break a successful rollout of 5G. As 5G demands higher bandwidth, lower latency, and other performance improvements over 4G LTE networks, operators like China Mobile look to the network edge and even the “on-premises edge” to meet these requirements.

The user plane, where the data itself passes through, is the block that must be moved to the edge. Control and configuration, on the other hand, can be kept in a central location. In order to make this work, China Mobile is asking vendors to work together so the UPF can be open and separated from the software control plane.

China Mobile further believes that FPGA hardware is the right programmable platform to handle the user data plane. They therefore called for delivery of an FPGA SmartNIC to offload the data plane from the server CPU, further speeding up and improving the performance of the network. When it comes to creating a universal platform to handle the user plane function, this can be accomplished best with an FPGA, because of the following advantages:

• Far better performance (throughput, latency) than standard CPUs, which are simply not built for network data flow processing
• Small footprint and low power consumption, ideal for the network edge
• Runs on standard off-the-shelf servers, where the embedded FPGA itself is a COTS platform, as opposed to the dedicated hardware of an ASIC
• Can be reprogrammed as needed to meet evolving standards and requirements
• Highly reliable with fast failover
• Enhanced security with IPSec tunnel endpoint aggregation handled inline such that data never needs to access the CPU

China Mobile’s initiative is an excellent example of the trend toward hardware disaggregation in the network, separating control and user functions for true network function virtualization, while keeping the infrastructure agile enough to adapt to ever-changing standards and requirements.

China Mobile’s open user plane function policy will allow more Tier 2 vendors to compete to supply 5G solutions, as the operators will be empowered to select the optimal solution for UPF rather than rely on the Tier 1 ASIC suppliers for entire systems. With China the clear leader in 5G planning so far, such an initiative should also encourage other network operators in China and around the world to follow China Mobile’s example. Ethernity looks forward to working with system integrators on creating a universal platform for 5G networking and taking an active role in future discussions toward developing Open UPF.

Ethernity Networks has natively separated the data plane from the control plane in FPGA-based acceleration solutions for over 15 years, deploying more than 600,000 systems and connecting more than 100 million end users.

In fact, Ethernity provided an earlier generation of this solution over EPC as early as 2015 that has remained in deployment since then. With our ACE-NIC100 SmartNIC and Router-on-NIC FPGA firmware, Ethernity offers an optimized, next-generation solution for 5G UPF acceleration that achieves the required high-performance benchmarks of 5G by offloading the entire data plane to the FPGA.

To link to the original article in Chinese about China Mobile’s initiative, click here.

By Louis Luo

by Barak Perlman

There is an increasing need for flexible cloud-based infrastructure and orchestration solutions to allow for dynamic placement of functions where and when they are needed. When it comes to the Telco edge, the gold standard for OpenStack-based virtualization software platforms is Wind River Titanium Cloud. Titanium Cloud is an ultrareliable deployment-ready cloud platform, capable of handling the rigorous demands of telecommunications and critical infrastructure networks.

Ethernity Networks and Wind River have worked closely together to show that Ethernity’s ACE-NIC100 FPGA SmartNIC can easily integrate with the Titanium Cloud virtualization framework. Titanium Cloud provides an edge-optimized all-in-one installation that runs an operating system with a hypervisor that enables the use of virtual machines (VMs) for network functions that are both lightweight and optimized for Telco.

The results: The ACE-NIC100 can accelerate any VNF on Titanium Cloud using Ethernity’s Router-on-NIC capabilities, with very little required effort toward integration. With the ACE-NIC using standard DPDK API calls for the ENET Flow Processor embedded on the FPGA, an Intel XL710 controller, and well-known i40e drivers, integration is seamless.

The ACE-NIC100 can be a separately managed entity. The ACE-NIC features the ENET Flow Processor, which runs on an FPGA on the ACE-NIC card and can be configured as a full router. This enables the ACE-NIC100 to provide both standard Intel-based Ethernet controller and router functions, achieving a true “Router-on-NIC.” As a Router-on-NIC, the ACE-NIC can provide a wide variety of Telco features through hardware implementation.  

The ENET Flow Processor is configured using the Ethernity SDK (Software Development Kit). The Ethernity SDK for ACE-NIC100 configuration is on-boarded on the Titanium Cloud controller, which is then used to configure policing, classification, and TAG/Tunnel editing. Two Linux virtual machines are instantiated with DPDK, each running the open-source TRex packet generator to emulate realistic traffic flow and provide per-stream statistics.

The integration demonstrates three distinct test configurations:

1. PCI-Passthrough + DPDK
   In this scenario, both of the ACE-NIC100’s 40GbE interfaces are configured with PCI-Passthrough, binding to DPDK drivers in the virtual machine.
2. SR-IOV + DPDK
   In this configuration, both of the ACE-NIC100’s 40GbE interfaces are configured with SR-IOV and a VLAN-based provider network, binding to DPDK drivers in the virtual machine. The traffic is run and validated twice – once at full rate, and once when the ACE-NIC100 is configured with per-flow rate-limiting policies using MEA CLI.
3. SR-IOV + DPDK + QoS
   In this scenario, one ACE-NIC100 40GbE interface is configured with SR-IOV, and the second 40GbE interface is configured as a data interface using Wind River’s Accelerated Virtual Switch (AVS). Wind River’s AVS ports connect a virtual machine to the AVS bound to the DPDK-AVP drivers in the VM. The traffic is validated twice, once at full rate and once when the ACE-NIC100 is configured with per-flow rate-limiting policies and VLAN-based provider networks defined for both interfaces.

The test emulates 40Gbps traffic with high diversity generated by multiple guest instances, connected to multiple virtual functions. The result is that the ACE-NIC100 easily couples with the Wind River AVS and enables cross-VM connectivity, while enforcing the rate-limit policy and other functions based on the virtual function traffic. In fact, the ACE-NIC100 is capable of fully offloading the VMs with high networking load to the FPGA, saving CPU cycles on Titanium Cloud.

The ACE-NIC introduces a router entity to the virtualization environment that allows Access Network traffic to enter Titanium Clouds’ virtual networks (for example, termination of PPPoE traffic or GTP tunnels), provides traffic management toward the external network, and adds VxLAN termination and translation to VLAN. And yet, the Router-on-NIC can be managed as a standalone managed entity, and the FPGA is transparent to the applications unless configured otherwise.

Moreover, by using SR-IOV for flows that do not run through the AVS, Ethernity has overcome the difficulty of supporting container-based virtualization. By instantiating multiple containers within a single VM, Ethernity was able to achieve per-container provisioning in hardware based on classification of the different flows arriving to the containers based on MAC address or VLAN, while applying per-flow policies.  Ethernity can enable provisioning per container by providing logic through the FPGA SmartNIC.

Furthermore, Ethernity was able to overcome the limitations of SR-IOV in terms of number of virtual functions. The FPGA SmartNIC can enable scalability to thousands of virtual functions (and therefore thousands of containers) within a single VM – classified, provisioned, and with policy application per-flow in the hardware.

“Service providers are seeking validated and market-ready end-to-end cloud solutions. To address this need, collaboration across the ecosystem is vital. We are working with innovators such as Ethernity Networks to create optimized, interoperable solutions for service providers and TEMs. By leveraging pre-validated virtual network elements, service providers can quickly achieve their goals such as reducing OPEX while accelerating the introduction of new high-value services,” noted Paul Miller, vice president of Telecommunications at Wind River.

By integrating Wind River AVS with the FPGA-based ACE-NIC100, Ethernity can deliver an approach capable of achieving efficient, scalable, high performance virtualized networking.

By working smoothly with Titanium Cloud, the ACE-NIC100 accelerates VNFs and adds routing functionality with very little required effort.

By Barak Perlman, CTO Ethernity Networks

Wind River is a global leader in delivering software for the intelligent edge. The company’s technology has been powering the safest, most secure devices in the world since 1981, and is found in more than 2 billion products. Wind River offers a comprehensive portfolio supported by world-class global professional services and support and a broad partner ecosystem. Wind River software and expertise are accelerating digital transformation of critical infrastructure systems that demand the highest levels of safety, security, performance, and reliability. To learn more, visit Wind River at www.windriver.com.

by Brian Klaff

At the annual DPDK Summit North America, Ethernity Networks CTO Barak Perlman was a featured speaker, providing insight into the role DPDK plays in 5G UPF hardware offload.

In fact, 5G was a topic mentioned in a couple presentations at the DPDK Summit this year. While few companies in attendance were as intimately aware of the trends within the telecommunications industry as Ethernity, a few speakers addressed edge virtualization, UPF offload, and/or GTP tunnels, all of which are integral components of a complete 5G deployment.

Barak’s presentation provided both an informative overview of UPF to those who were less familiar and insightful ideas on how UPF offload can be achieved for those who work with FPGA SmartNICs. The main discussion centered around using DPDK hardware offload APIs as the interface between the UPF software and the SmartNIC. This seems to be a natural choice for DPDK users, and yet, this is the first time the topic was formally presented.

Perhaps most relevant to the DPDK community were Barak’s suggested improvements to rte_flow in order to facilitate full UPF offload. Rte_flow was another hot topic at the Summit, with Intel, Mellanox, and Microsoft addressing the benefits and challenges of working with this hardware offload API. Intel presented a good overview of the rte_flow library and discussed the practical implications of using rte_flow for partial or full offload of OVS traffic.

Microsoft brought significant weight to the discussion as an rte_flow user in its cloud environments, and highlighted the fact that rte_flow is the de facto standard for hardware offload, supported by numerous NIC vendors. However, Microsoft spoke of their concern that some APIs might behave differently with different vendors, and suggested ideas for testing and resolving this issue.

Mellanox also spoke of rte_flow in relation to enabling hairpin offloading, in which data never needs to reach the CPU, but rather is entirely processed within the NIC. This was validation of the method that Ethernity suggested in its UPF presentation, offering the same packets-per-second performance with zero CPU involvement compared to the many cores required by partial offload. This allows organizations to Stop Burning CPU Cores!

One more hot topic at the conference was containers, which are rapidly supplanting virtual machines as the go-to technology for embedding applications in a virtual environment. Multiple sessions were devoted to applying DPDK to container applications and scenarios, as well as DPDK’s interplay with Kubernetes, the open source system for orchestrating and managing containers.

Overall, the DPDK Summit was an opportunity for some of the primary companies in the open-source networking world to share their experience and expertise toward improving this primary set of API libraries and drivers for network processing. Ethernity was happy to take an active role in these discussions

by Eugene Zetserov

Broadband World Forum has traditionally been an event for fixed networks, but in recent years it has addressed the needs of fixed, mobile, and cable networks, covering a variety of topics, including 5G, virtualization, open source, NG PON, DOCSIS, and edge computing. BBWF2019, earlier this month in Amsterdam, was all about providing reliable connectivity and delivering new high-speed services in broadband access networks that leverage a variety of technologies, virtualization, and open source principles.

5G and new services that require ultra-low latency are totally changing the required network infrastructure, demanding gigabit connectivity and 100% coverage, which are not so simple to achieve. The paradigm of the Network Edge has been conceptualized as a solution for implementing IoT and handling the special requirements of high-resolution video content and low-latency applications, such as interactive gaming, autonomous cars, and AR/VR. But the Edge differs from the core in its physical infrastructure, its ownership, and in the necessary tasks to fulfill its connectivity and service requirements. At the same time, the Edge must be a natural extension of the core network, a so-called vector from the perspective of management and operations. It needs to run applications migrated from the core and provide the relevant resources despite restricted capabilities. This means it must be high throughput on the one hand and low latency and low power on the other.

To this end, Broadband Forum and Open Networking Foundation announced at the start of the 2019 event in Amsterdam a new agreement to pair some of their projects that provide abstraction of broadband access, namely Open Broadband-Broadband Access Abstraction (OB-BAA), SDN Enabled Broadband Access (SEBA), and Virtual OLT Hardware Abstraction (VOLTHA), helping operators who are looking to interconnect different parts of their network with open source solutions and systems from various suppliers.

A primary example was showcased by ONF at BBWF2019. ONF presented its SEBA platform for vBNG, which supports a multitude of virtualized access technologies at the edge of the carrier network. SEBA integrates software components and common APIs for handling interoperability, together with acceleration products, including SDN switches and SmartNICs that support separation of the control and user planes (CUPS). The SmartNICs provide the necessary acceleration at the Edge, where compute and power resources are especially limited, but high performance and low latency are still required.

ONF, of course, was founded under the premise that telecom operators want to avoid any vendor lock-in, not only of system vendors, but also chipmakers. The market has evolved to demand open solutions not only for software, but also for acceleration, and the FPGA is the top option for providing a solution. There are numerous FPGA flavors, ranging from low-throughput models up to versions that offer hundreds of gigabits of throughput with integrated ARM and DDR for a simplified, optimized data processing experience. FPGAs are programmable, even once field-deployed, meaning they are future-proof, and they are becoming much more accessible with newly available tools to enable the community to contribute to further FPGA development. This is exactly where Ethernity Networks enters the picture by offering FPGA-based acceleration solutions to the portfolio of the Network Edge.

ONF is now planning to show a 5G User Plane Functionality (UPF) SEBA platform at Mobile World Congress in February. This should be of particular interest to operators and system integrators.

Ethernity is ready with its 5G UPF Acceleration solution, which leverages our patented ENET Flow Processor technology along with standard DPDK APIs to offload the data plane to an FPGA-based SmartNIC. Our ACE-NIC100  integrates easily with third-party UPF software networking elements from any vendor to offload user plane data, thereby releasing server CPU cores, enhancing scalability, assuring deterministic performance, improving latency, and providing future-ready programmability. This accelerates the entire 5G network with the lowest possible TCO.

As both telecom operators and cloud providers move their computing resources to the network edge, to bring applications closer to users and reduce latency, the question has arisen as to who the dominant players will be.

Whereas the trend for the past decade or so has been for the big three cloud companies to control much of the technological advancement and revenue streams related to cloud services, the move to edge computing is suggesting a shift of that industry trend toward the telecom operators.

Since operators already own so much existing infrastructure at the network edge, the pendulum is swinging in their direction. We are seeing operators turning existing central offices into advanced, next generation versions that serve as new mini-data centers nearer to the biggest sources of required resources.

You can be sure the big three cloud companies – Amazon, Microsoft, and Google – are concerned. They have built truly impressive centralized data centers, which still have tremendous value, but which are just too far away in a world that is becoming increasingly concerned with communication latency. Latency is a critical issue in the age of IoT and real-time healthcare, manufacturing, and other applications.

Industry analyst Chetan Sharma in a recent Bloomberg article had this to say: “Over time, cloud will be primarily used for storage and running longer computational models, while most of the processing of data and AI inference will take place at the edge.” Sharma sees a huge edge market in the next decade – worth more than $4 trillion by 2030.

And as we noted, who dominates the edge? It is the operators and owners of cellular towers, who control the valuable real estate that comprises today’s edge, and these assets will only become more precious in the coming years.

Sharma contends that the big cloud players are realizing that it’s in their best interests to partner with operators in order to get access to the edge. But again, that reinforces the controlling position that operators could hold going forward.

We have talked a lot about the edge, both here and in our white paper, Enabling the Virtualized Edge with SmartNIC Data Acceleration. The white paper addresses important issues about the edge, including ways that operators can meet the huge demand for delivering virtualized services efficiently given the scarcity of space and power at the network edge. It also touches on the need for the cloud providers to extend their cloud networks closer to the edge to benefit from the low latency provided by being physically close to end users.

Because edge sites are – and will be – compact, with a reduced physical and power footprint, operators are going to need to provide maximum levels of computing, networking, and security in a small space. That is where FPGA-based SmartNICs are the ideal solution, as they are much more space-conscious and power-efficient than adding servers, and they are optimized for networking and security functions, freeing CPUs to handle the control functions and user applications for which they were intended.

by David Levi

Residents of technologically advanced countries sometimes take for granted the blessings of instantaneous access to the wealth of information on the internet, the almost infinite options for media and entertainment, and connections to friends, family, and colleagues anywhere. All those advantages are only getting better with the advent of 5G broadband. With 5G will come interconnectivity between devices, the Internet of Things (IoT), augmented/virtual learning and gaming, and other advanced latency-sensitive applications.

But what about the developing world? 5G networks’ potential to replace physical fiber networks and better penetrate areas currently lacking a real broadband infrastructure could transform these regions economically.

Broadband is clearly a driver for economic growth. A World Bank report calculated that for every 10 percent that a developing economy can improve broadband penetration, it benefits from a nearly 1.4 percent improvement in GDP.

This advancement creates new jobs in cities and rural areas. Existing businesses gain access to innovations and new business models. Small businesses can grow and even become globally competitive. The quality, skills, and technological savvy of local workforces improve.

5G broadband can also improve education, health care, and emergency services. The ultra-low latency anticipated with 5G will enable instantaneous communications for medical imaging, monitoring, and screening, and make remote surgery possible. With 5G-connected IoT sensors, agricultural growth can be optimized through better water and fertilizer management, reducing the risks of droughts and agricultural crises and aiding the environment.

But there are challenges to bringing 5G broadband to the developing world. First is a lack of infrastructure; there is little in the way of cell towers and base stations, and much of the equipment currently deployed is too antiquated to ever support 4G. We need a significant leap – essentially skipping a generation of technology – but this requires a large investment and some hard choices.

Financial realities inhibit service providers when it comes to investing in rural and underdeveloped areas. In the face of the high costs of network deployment and management, plus a generally low level of revenue per user, providers must keep costs as low as possible. And with concerns about the reliability and availability of power in many developing areas, a low-power solution is an absolute requirement for 5G broadband projects.

If service providers focus on deploying distributed compute resources closer to the network’s edge instead of laying fiber, they will be more likely to be able to support the high bandwidth low latency services that will most benefit the developing world. The key is to move from a heavy network infrastructure to the network of the future – one that is agile, virtualized, and cloud-based, with smaller and more efficient virtual central offices strategically located at the edge of the network.

Network acceleration is critical for reducing latency and minimizing costs associated with large rural 5G deployments. Such solutions can save space and power, especially when the network acceleration is performed with hardware that can replace traditional server cores. For example, by using FPGAs embedded in network adapters – SmartNICs – to handle all data plane functions, operators can significantly reduce the need for servers, saving both space and power.

Programmable for a wide range of functions, FPGAs offer deterministic, low latency performance and low power and space requirements, along with high flexibility and long-term cost-effectiveness for carriers. This enables easy future upgrades compared with other solutions.

Ultimately, to achieve the great benefits of 5G broadband, FPGA-based network acceleration will be a key element in reducing power, space, and cost. It will further enable the high bandwidth and low latency needed for delay-sensitive services – the kinds of services that can positively transform remote rural communities and underprivileged urban centers.

by Shavit Baruch

With the imminent spread of 5G and IoT as well as greater demands on enterprise WANs, there is a need for network edge platforms that provide efficient and secure high-speed connectivity.

That is how Roy Chua, principal at the research and analyst firm AvidThink, describes the situation. And that is the rationale behind our new ENET Universal Edge Platform, which we introduced last week.

The ENET UEP is optimized for network edge applications. It offers high performance, solid security, and nearly unlimited flexibility in its protocol and port configurations. Those are critical needs at the edge, where service providers need to be able to do more in less space and with lower power consumption.

That is true for whatever form the network edge takes, whether that is a cell tower, a street cabinet, a multi-dwelling distribution/cable box, or a remote business office.

The ENET UEP is an edge-optimized, compact, low-power, and FPGA-based (so it is highly programmable) network appliance. On top of its 40 Gbps of networking capacity and 10 Gbps of IPSec security, its modular design makes it easily adaptable for use cases such as:

• High-end network interface device (NID) for demarcation of the WAN from the LAN networks. With its dual-core ARM processors, the ENET UEP can handle all control functions, while the onboard FPGA handles the data path.
• Mobile backhaul with XGS-PON. Because it is compact and power-conscious, the ENET UEP can be located at cellular base stations to provide cell site aggregation and XGS-PON connectivity with the optical line termination.
• Distribution point unit or multi-dwelling unit. For DPU or MDU applications, the ENET UEP can be converted via its modularity to handle G.fast, and in addition it can offer cascading switching for even greater distribution capacity.
• Internet of Things. The ENET UEP can support IoT aggregation elements, such as a radio modem for the IoT sensor network.

One major differentiator for the ENET UEP is that it offers a unique PCIe connection to any standard server, which enables it to be used for NFVI acceleration. Analyst Roy Chua drew attention to that, noting that the flexibility afforded by the PCIe connection to a standard server allows the ENET UEP to act as a viable accelerator to NFV workloads at the network edge.

More information about the ENET UEP is available here.

At the 2019 edition of the Edge Computing Congress last week, Ethernity Networks was in attendance to promote our 5G UPF Acceleration solution and Universal Edge Platform network appliance. The conference had been held in Berlin the previous two years, but it moved to London this year, with an impressive agenda of top speakers and a small but power-packed exhibition area.

The most prevalent topic of conversation was, of course, Multi-access Edge Computing (MEC), with emphasis on both the required infrastructure to enable the edge and the development of the applications that will benefit from edge computing. The clear takeaway from the two days of sessions was that operators are ready to invest heavily in the edges of their networks.

There are various reasons for this, but first and foremost among them is the push to implement 5G. There is no doubt that 5G is coming, with initial trials already underway and widespread rollouts expected in 2020. But from the analysts that Ethernity spoke to, the overarching sense is that operators have both high expectations and serious flaws in the planning. For example, many 5G deployments are barreling ahead and relying heavily on virtualization without considering whether the existing hardware is able to support such a network. In one presentation, Julian Bright of Ovum emphasized the need for flexible, scalable hardware that is capable of being hosted in any location, from cell sites to remote closets.

Of course, this plays perfectly into Ethernity’s strength. Our ACE-NIC100 FPGA SmartNIC is ideal for deployment at the edge, saving customers precious space and power, while providing high performance that can easily scale to handle even the most demanding access networks.

Despite the commitment to spend on edge computing, a number of major concerns still exist.  For example, given the need to build hundreds of edge facilities, operators worry about the total cost of ownership, including maintaining a workforce across the edge. Operators therefore need networking automation at the edge, and Ethernity’s flow processors can help support such automation with programmable OAM features on the FPGAs.

Another issue that operators face at the edge is that of standards, regulation and SLAs. The ETSI MEC Industry Specification Group is working on yet another phase of specifications and edge computing use cases for telecom companies that, unlike their cloud provider counterparts, are heavily regulated and forced to maintain high quality of service experience. Ethernity’s 15 years of networking experience and our support for carrier-grade SLAs and hierarchical quality of service on our ACE-NICs make us an ideal partner for operators with edge ambitions.

A couple of other questions regarding 5G permeated the event:

• Is the race to implement 5G-capable infrastructure driving developments in edge computing, or are developments in low latency-dependent applications, such as Industrial IoT, autonomous vehicles, augmented/virtual reality, and gaming the stronger drivers for progress at the edge?
• Are the 5G trials and initial deployments realistic indicators of success, or are the current bandwidths still too low to predict success upon full rollout?

Ethernity is confident that the ideal 5G deployment will combine network virtualization with hardware acceleration.  By offloading the 5G User Plane Function to an FPGA SmartNIC, the server CPUs can be dedicated to the functions they are better at handling, namely control function and user applications.

FPGAs are better optimized for data transport than CPUs, and Ethernity’s patented ability to reduce the space required for flow processor logic allows our FPGA SmartNIC to include full router capability directly on the NIC. This unlocks the potential for tremendous savings in space, power, and CPU utilization for cost-conscious telecom operators.

After many positive conversations with industry analysts and potential partners and customers, we are already looking forward to the 2020 Edge Computing Congress. By then we expect to know a lot more about 5G progress and the need for Ethernity’s solutions.

When it comes to 5G, the User Plane Function (UPF) is an essential element of the evolving network development. The CUPS (Control and User Plane Separation) architecture put forth by 3GPP separates the Serving Gateway and PDN Gateway functions into control and data plane components, and the UPF serves as the data path. This approach enables packet processing and traffic aggregation to be handled closer to the network edge. That is where service providers are so focused today as they seek to increase bandwidth efficiencies, reduce latency, and enable service-enhancing capabilities such as network slicing. In this new architectural model, the role that 5G UPF acceleration plays in increasing network efficiencies is receiving a lot of attention, and deservedly so.

For us, at Ethernity, the separation of the control and data planes is native. The fact that 3GPP 5G architecture has addressed this in Release 14, 2017 and operators are starting to implement it today only serves to prove our concept. So how do we handle it? The Ethernity 5G UPF Acceleration solution is to leverage our patented ENET Flow Processor technology along with standard DPDK APIs to offload the data plane to an FPGA-based card. This allows acceleration of the entire 5G network with the lowest possible TCO. We discuss this approach at length in our 5G UPF Acceleration Solution Brief.

Briefly, what we do is use our FPGA-based ACE-NIC SmartNIC, which integrates easily with third-party UPF software networking elements from any vendor, to accomplish several things:

• Offload user plane data, to release server CPU cores
• Enhance scalability
• Assure deterministic performance and improve latency
• Provide future-ready programmability

3GPP 5G architecture supports disaggregation by enabling the UPF to be placed at the network edge, closer to end-user locations. This allows a service provider to achieve better performance, reduce networking overhead, and lower costs. With a small footprint and low power requirements, the 5G UPF Acceleration solution is truly optimized for edge deployment. In this case, power requirements are extremely important, so our SmartNIC is the perfect product to address the issue, with a super-efficient ratio of performance per Watt.

There is additional support for 5G Quality of Experience, including high bandwidth, low latency, dense connectivity, and multitenancy. The easily programmable data path can quickly adapt to a service provider’s particular requirements and evolve with the ever-changing architectural requirements of the mobile market. Attempting to achieve high data rates with a standard general CPU alone can lead to data loss, packet count loss, and data starvation, so our FPGA-based ENET Flow Processor is an ideal complementary tool to assist x86.

The solution can be implemented via an Ethernity ACE-NIC for accelerating 5G UPF as a virtual network function and for accelerating the NFV infrastructure (NFVI) based on DPDK.

UPF is foundational to a new generation of service-based architectures. As service providers increasingly focus on the network edge and seek to gain efficiencies there, the ability to accelerate UPF through a FPGA-based SmartNIC solution will prove to be a highly valuable tool in their toolbox.

by Lior Mishan

The evolution of central offices into small cloud data centers – with commodity servers and network function virtualization infrastructure (NFVI) replacing fixed-function hardware devices – is the subject of the new, in-depth 2019 Next Gen Central Office report by the research firm AvidThink.

These Next Gen Central Offices, are needed so service providers can prepare, as the report says, “for an onslaught of high-speed devices and bandwidth-hogging content.” Providers are facing the limits of inflexible, hardware-based centralization and the need to distribute workloads to be closer to the user, as well as to meet the clear industry demand for lower latency performance.

5G will help, but the bold promises of ultra-high speeds and millisecond latency can only be achieved by offloading backbone networks and moving content and services to the edge.

We have written about this important topic ourselves, in our white paper, Enabling the Virtualized Edge with SmartNIC Data cceleration, and in several articles in industry publications, the most recent one by CTO Barak Perlman in Lightwave.

As the report notes, “The virtualization of carrier networks from fixed function hardware to SDN and NFV has improved the efficiency and capacity utilization of carrier infrastructure and their ability to rapidly introduce new services. These improvements have also increased customer expectations and the need for service innovation and increased speed and flexibility in service delivery.”

In addition, a software-defined infrastructure also facilitates moving services to next-generation platforms like containers, serverless functions, and edge data centers to create competitive differentiation.

“A software-defined Next Gen Central Offices has strategic value by providing the foundation for secure multi-tenant services that enable new business models such as the ability to more easily resell capacity and services to third-party service providers; provide service bundles that include third-party products; and allow customer self-service installation and management of select network services,” the report explains.

It goes on to emphasize that  “A software-defined infrastructure that can span both massive scale out Central Offices and small edge locations offers carriers and other network service providers significant financial, technical, operational and competitive benefits.”

These benefits include:

• Faster time to market for new services and updates.
• Greater scalability as virtual services displace physical hardware to allow rapid expansion of service capacity in response to short-term spikes in demand.
• Reduced long-term CapEx and OpEx.
• Reduced traffic and congestion in the network core as edge Central Offices offload high-demand content and services, enabling lower latency for service and content delivery.
• Likely improvements in system-wide energy efficiency with fewer network hops and fewer pieces of equipment needed for service delivery.
• Architectural decoupling of data and control planes, which leads to decoupling of service delivery from specific hardware, allowing independently scaling capacity and services/content.
• Service customization flexibility, including the ability to deliver customized services and applications from a microcloud aimed at particular customers and with contextualized services and resource allocations tailored to them.
• More choice of hardware and software products and suppliers thanks to the use of standard commodity hardware and OSS, which also translates to fewer technical support challenges since many components are interchangeable.

This must-read report goes on to explore in depth the characteristics and technologies associated with Next Gen Central Offices – including the role of FPGA-based SmartNICs. It offers a series of recommendations and next steps for service providers and enterprises. The report can be downloaded here.