Making Network Disaggregation a Reality

Benefits of Disaggregation in General

Disaggregation has enabled the move toward agile networks. Whereas once the standard was for networks to rely on ASIC (application-specific integrated circuit)-based monolithic appliances that bundled proprietary software into a vendor-locked device and integrated only with other offerings from that vendor, disaggregation has changed the paradigm to overcome such limitations.
Thanks to disaggregation, software-based functions run on top of CPUs inside standard servers from any of a number of different vendors, and open stacks are used to communicate between virtual machines for application and service chaining.
The concept of disaggregation has enabled enterprises to eliminate the cost associated with procurement and maintenance of hardware equipment. By deploying these solutions, they can also handle multiple workloads with maximum uptime and improved performance.
Disaggregation offers much greater agility in addressing evolving requirements, provides added control over both current and future functionality, and eliminates software vendor lock-in. It can therefore generate significant savings in a company’s operational expenses.

Disaggregation at the Network Edge

While disaggregation has become standard in data centers and cloud networks, it has yet to penetrate fully at the network edge.
At the network edge, operators have gained some measure of choice in their software vendors and greater configurability of the control plane, but data plane functionality within carrier-grade white box switching and routing hardware is still limited by the ASIC-based silicon therein.
Open platforms offer a choice of operating system or control software only, while maintaining vendor lock-in when it comes to the underlying hardware-implemented data plane. Moreover, that data plane functionality often lacks many of the advanced features that are required by operators at the network edge.
Standards and requirements of the network edge are still evolving, such that operators have to consider agility and futureproofing when selecting edge switching hardware. ASIC-based switch hardware, even in white box devices, cannot be upgraded to address new data plane functionality requirements.
The issue is that no standard platform has yet emerged to enable full disaggregation of the data plane hardware. No standard has yet become the de facto generic platform for data plane functionality and feature development.

FPGA: The Platform for Network Disaggregation

One technology that is already being used to a large degree for accelerating the data plane in the data center market is the Field Programmable Gate Array (FPGA). FPGAs are programmable hardware that are optimized to handle networking and security traffic, which can be incorporated in a white box edge device by way of a system-on-chip (SoC) or smart network interface card (NIC).
Many networking and security functions are highly CPU-intensive, burning through CPU cores with varying levels of performance. When the data functions are offloaded from CPUs to FPGAs, performance not only improves; it also stabilizes and is more deterministic, and it saves CPU cores for the compute and control functions they were meant to handle.
FPGAs offer complete network hardware disaggregation. An operator can select the required FPGA just as they would select the required CPU and processing power in a server. FPGAs are available from multiple vendors, all of which offer toolkits to enable development. Operators would also have the freedom to choose the code that runs on an FPGA-based network interface card, selecting from an array of firmware vendors.
FPGA-based NICs can address a wide range of acceleration use cases and can be redesigned very quickly, with new functionality added on top of the existing IP. They are programmable and reconfigurable so the same hardware can be repurposed to handle new applications.
FPGAs are also deeply pipelined, with massive on-chip memory and external memory to DRAM. They can therefore deterministically provide high throughput with extremely low latency and jitter. FPGAs offer ASIC-like performance with the programmability of software.
As a result, FPGAs are the ideal platform for truly disaggregating hardware at the network edge in white box edge switch/routers and OLTs. FPGAs improve performance while avoiding vendor lock-in, and they futureproof the network, thereby saving on long-term operating expenses and reducing total cost of ownership.
By establishing FPGAs as the optimal open standard platform for handling the data plane in edge devices, operators gain the flexibility to overcome vendor lock-in and the programmability to futureproof their networks, which is the ultimate requirement for today’s network edge.

Ethernity’s Value

Ethernity Networks brings over 15 years of experience in FPGA programming for telecommunications applications. Our family of ACE-NIC FPGA SmartNICs offers the high bandwidth and low latency that are necessary to accelerate user applications and 5G deployments at the network edge, while slotting into existing servers, thereby reducing the need for additional boxes and saving space.
Ethernity’s patented technology also reduces required programmable logic by up to 80%, enabling the use of smaller FPGAs, for more affordable acceleration solutions. Our FPGA-based ACE-NICs address the security requirements of the edge as well, enabling network isolation and user segregation to prevent attacks on edge sites and user devices, as well as offering IPSec VPNs and tunnel termination.
Ethernity Networks has also proposed the UEP-200, a multiservice edge compute platform with open and disaggregated FPGA-based switch/router hardware that offers the utmost in networking flexibility. Within a single box, the UEP-200 can connect 10G/25G Ethernet, xPON, fronthaul mobile access, remote cable modem deployment, and 5G User Plane Functionality.
The UEP-200’s distributed switch architecture can incorporate multiple FPGAs within the same box to enable growing to 600Gbps, while it can also serve as a virtual chassis by cascading units to build an aggregation cluster of 1Tbps. The UEP-200 is the ultimate realization of the vision of complete network disaggregation.