Thursday, November 11, 2021

Blueprint: Reimagining Metro Networks for Edge Computing and Beyond

by Julius Francis, Sr. Director, Product Marketing, Juniper Networks

As new advancements are made in 5G, Internet of Things (IoT) and edge computing, doors are opening for network operators to play a more prominent role in the delivery of next generation consumer and enterprise services. However, these opportunities come paired with new challenges. Specifically, these new services and requirements drive increased traffic and diverse traffic patterns throughout networks, demands that will have the most significant impact on metro networks. 

Metro networks will play an increasingly critical role in service delivery innovations; where all services accessing the network converge (e.g., mobile devices, laptops, gaming consoles, IoT devices and sensors, etc.). Traditionally, metro networks were built to funnel traffic into centralized edge and core resources – a characteristic that is rapidly evolving as edge service technologies are increasingly virtualized and instantiated across new cloud resources. The result is more efficient use of metro facilities as more and more workloads remain in the metro reducing latency and improving user experiences. As a result, metro traffic is projected to grow four times faster than anywhere else in the network by 2025.  

To leverage this opportunity, network operators must rethink how metro networks are architected to unleash new digital experiences – from how they’re designed, to how they’re used, to how their resources are allocated. And operators must fully embrace automation to make this new level of orchestration possible. Focusing on this transformation will help operators deliver better user experiences, helping them play a more prominent and profitable role in the digital ecosystem.

Drivers of Change

Over the next five years network traffic growth will continue, driven largely by the increased adoption, virtualization and prominence of edge applications. Cloud computing resources are increasingly being deployed across cloud resources in parallel with the metro, moving the execution of edge services from regional data centers to the cloud – a trend that is expected to increase over the next decade. Many next-generation edge applications and services require low latency and customized SLAs, which legacy networks based on regional data centers may struggle to provide. 

Further, the proliferation of high-quality and mobile video content will increase demands on metro networks. Video content will increasingly be cached within the metro and distributed to users locally rather than from central cache locations within the core. 

As new distributed 5G and edge services push more workloads out to distributed cloud resources, existing metro networks built for conventional networking will no longer be sufficient to leverage the benefits associated with cloud-based edge computing. All of these issues are driving the need for change in network architecture, infrastructure and operation. 

Reimaging Metro Network Design

Many legacy metro networks are designed using ring topologies, which enable metro networks to carry traffic from access nodes, across an aggregation network, to regional data centers. While there are many instances where the ring design can solely provide adequate performance, operators are turning to a new industry trend to achieve the flexibility and agility needed to handle increasing network traffic and fully leverage the advantages of edge computing – the ring topology paired with spine-leaf. In a spine-leaf topology, each network element has a dedicated connection point, which reduces the transit traffic and throughput requirements on each device. Therefore, this combination provides more scalable, flexible and resilient bandwidth and allows metro networks to handle higher rates of traffic and bandwidth needs while leveraging the power of two topologies. 

By reimagining their metro network design, operators can future proof their organizations for the challenges of the 5G, IoT and cloud era. To attain a next-generation metro network, network operators must stay laser focused on achieving the following:

  • New network architectures. To keep pace with shifts in metro network requirements and support the dramatic growth in traffic, metro topologies can leverage ring topologies, spine-leaf topologies, or a hybrid of both networking architectures.
  • Scalability and flexibility. To achieve elastic scalability, new routers and network architectures with scalable bandwidth, real-time service monitoring and control, and end-to-end automation must be implemented.
  • Service and application intelligence. Next-generation metro networks leverage network slicing to intelligently steer traffic to the right physical and virtualized resource by identifying the most efficient path across the network to optimize user experience. Future-proofed metro networks must have cloud computing levels of orchestration and automation to keep pace with evolving services, applications and user expectations.
  • End-to-end automation. Network automation is necessary to reduce complexity and help engineers manage the increased traffic load and simplify operations. Manual operations are no longer capable of keeping pace with the dynamics of a next-generation metro network.
  • Converged multiservice metro networks. To manage the increasing scope, scale and complexity onto one common network, next-generation metro networks will need to eliminate siloes from legacy networks to focus less on process and more on end-user outcomes.

Metro Networks for Next-Gen 5G, Edge and IoT Services 

The 5G, IoT and cloud era brings incredible opportunity in every market vertical, but there will be growing pains along the way. The key to addressing new bandwidth, latency and orchestration challenges will be for operators to advance and modernize their network architecture. A metro network designed for yesterday’s static transport and aggregation needs simply won’t be capable of supporting the service delivery needs of 2021 and beyond.

By incorporating architectures built for automation and scale, metro networks that offer flexible network slicing, service-aware technologies and cloud-scale capabilities, will exceed user expectations for every service they deliver—while reducing the complexity of operations.

ADVA positions FSP 3000 OpenFabric1200 for 400G metro aggregation

ADVA introduced its FSP 3000 OpenFabric1200 for 400G metro aggregation.

The FSP 3000 OpenFabric Series combines open 400 Gbps pluggable coherent technology with OTN switching on a blade, enabling client services to be groomed into 400 Gbps wavelengths. The platform features 1200 Gbps interface capacity and supports Ethernet and OTN services from 10 Gbps to 400 Gbps.

ADVA says the open switchponder card can be added to any network without requiring additional adaptors. In addition, the FSP 3000 OpenFabric1200 can work in multiple operation modes: ADM on a blade, DWDM muxponders, or OTN cross-connect. 

“The strain on today’s transport infrastructure is growing like never before. Today’s operators need new solutions that can meet rapidly rising data demand while also driving down costs. That’s why we’ve developed our FSP 3000 OpenFabric1200. It offers a simple way to dramatically increase fiber efficiency by supporting lower-speed service grooming at the network edge. By deploying OpenFabric1200 in ring or hub-and-spoke topologies, operators can significantly reduce switching and cross-connect requirements in the core,” said Christoph Glingener, CTO, ADVA. “What’s more, our open Terabit-scale switchponder solution can be easily added to any network.”

Marvell shipping SONiC-enabled switch silicon to tier-1 clouds

Marvell confirmed that it is shipping SONiC-enabled production switch silicon of its Teralynx 9K and Prestera 8K devices in high-volume to tier-1 cloud customers. 

Marvell's Teralynx and Prestera devices support Switch Abstraction Interface (SAI) and Software for Open Networking in the Cloud (SONiC), which is an open-source network operating system (NOS) based on Linux. 

Marvell says SONiC running on its Teralynx and Prestera switch silicon brings greater flexibility and choice to cloud data centers grappling with the demand for greater bandwidth as data growth continues to surge.

"Marvell remains deeply committed to open systems and will continue to invest in and support SONiC solutions for our customers," said Guy Azrad, senior vice president and general manager, Switch Business Unit at Marvell. "Marvell's high-volume shipping of SONiC-enabled production switch silicon to tier-1 cloud providers showcases our technology leadership in combining our advanced packet processing architecture with the power of an agile, open network operating system."

"We have deployed Teralynx based switches in production across multiple data centers with the SONiC network OS. These switches deliver high performance, low-latency and rich telemetry required to run modern applications in our infrastructure," said Shawn Zandi, head of Network Engineering at LinkedIn. "With Innovium's acquisition by Marvell, we look forward to benefiting from greater investment into the Teralynx roadmap and Marvell's continued commitment to SONiC."

"Celestica delivers open disaggregated solutions that offer customers flexibility and compelling total cost of ownership," said Gavin Cato, VP, Celestica Hardware Platform Solutions. "We have been collaborating with Marvell to build high performance switch systems that are now deployed at scale by top cloud service providers, addressing the ever increasing need for bandwidth."

"Edgecore looks forward to the continued collaboration with Marvell to bring leading networking solutions for our service provider, data center, and enterprise customers," said Heimdall Siao, president of Edgecore Networks. "With Marvell's technology and Marvell's proactive support of the open SONiC software ecosystem, Edgecore is able to provide high performance and cost-effective solutions using community based open networking solutions which allows our customers and partners to achieve more."

Credo pushes LP SWITCH Active Electrical Cable to 200G

Credo announced the 200G HiWire LP SWITCH Active Electrical Cable (AEC) for connecting servers to to two Top-of-Rack (TOR) switches for hitless failover. The solution architecture was jointly announced in July 2021 by Credo and Microsoft, whose SONIC Dual TOR Management Container controls the SWITCH AEC.

The HiWire LP SWITCH AEC 200 has QSFP56 (Quad Small Form Factor Pluggable) connectors with integrated Credo PCS termination. The cables support 50G, 100G, and 200G with optional speed-shifting between 56G PAM4 and 28G NRZ lanes on both the NIC and TOR ends. The new 200G AECs join Credo’s diverse HiWire family for 100G, 200G, 400G, and 800G systems.

“Thanks to Credo’s advanced purpose-built SerDes architecture, our 200G LP SWITCH AEC has doubled its speed in the last six months while staying inside the same power envelope.”

Data processing and storage power can be monetized but not the power consumed for data transport. Credo’s low-power LP SWITCH AECs help minimize data transport power and enable data center architectures that eliminate redundant failover server racks.

“Power is the most important and most difficult parameter to control in hyperscale networking,” said Don Barnetson, Vice President of Product, AECs at Credo. “Thanks to Credo’s advanced purpose-built SerDes architecture, our 200G LP SWITCH AEC has doubled its speed in the last six months while staying inside the same power envelope.”

“As NIC speeds increase rapidly in the hyperscale datacenter, reliability must not be compromised,” said Lihua Yuan, Partner Development Manager at Microsoft. “Credo’s 50G and 100G HiWire LP SWITCH AECs are already proven and up-streamed in SONiC with the 200G LP SWITCH AEC creating a path for next-generation NIC speeds in 2022.”

“Emerging trends such as AI and 5G are driving exponential growth in the performance demands of the modern data center,” said David Iles, Senior Director of Ethernet Switches at NVIDIA. “The NVIDIA Networking platform equips innovators as Credo with the network bandwidth and low latency required to supercharge the modern data center with breakthroughs such as the HiWire LP SWITCH AECs.”

"Reliability is more critical for telecommunication service providers than for hyperscalers," said Keyur Patel, Founder and CTO, Arrcus. “Credo’s HiWire LP SWITCH AECs provide 5x9s (99.999%) rack reliability and enable us to ensure autonomous high availability (HA) from the Arrcus Connected Edge (ACE) platform with a sub-microsecond failover."

SFP-DD MSA issues two specs

The Small Form Factor Pluggable Double Density (SFP-DD) Multi-Source Agreement (MSA) Group has released updated 5.0 hardware specifications and drawings for the SFP-DD, SFP-DD112, and SFP112 pluggable modules. 

This revision enables SFP112 operating at 112 Gbps and SFP-DD112 operating at an aggregate rate of 224 Gbps; aligning with next generation higher speed networking, storage and access equipment. This new revision also includes an SFP112 module specification to ensure the industry roadmap from SFP28/SFP56 to SFP112 and SFP-DD to SFP-DD112 maintain overall compatibility.

SFP-DD MSA revision 5.0 hardware specification includes significant signal integrity enhancements to address 112 Gbps differential signaling. Three new clauses have been added as follows: Chapter 5 for SFP112 electrical and management interface requirements, Chapter 8 for SFP-DD 112G mechanical and board definition, and Chapter 9 for SFP112 mechanical and board definition. In addition, the new revision includes an ePPS/Clock signal definition for SFP-DD/SFP-DD112. Lastly, TS-1000 Normative Module and Connector performance requirements for SFP112 were added into Appendix A.

SFP-DD MSA promoters include Alibaba Group, Broadcom, Cisco, Dell Technologies, Hewlett Packard Enterprises, Huawei, II-VI Incorporated, Intel, Juniper Networks, Lumentum, Molex, Nvidia and TE Connectivity. Contributors include Accelink, Amphenol, AOI, Eoptolink, Foxconn Interconnect Technology, Fourte International, Genesis Connected Solutions, Hisense Broadband, Infinera, InnoLight, Maxim Integrated, Multilane, Nokia, Senko, Source Photonics, US Conec, Yamaichi Electronics, and ZTE.

FCC announces over $700M in RDOF funding in 26 States

The FCC is ready to authorize $709,060,159 in its fourth round of funding for new broadband deployments through the Rural Digital Opportunity Fund.  

The 26 states slated for funding include Alabama, Arizona, California, Florida, Georgia, Illinois, Indiana, Iowa, Kentucky, Louisiana, Michigan, Minnesota, Mississippi, Missouri, Montana, New Hampshire, New York, North Carolina, North Dakota, Oregon, South Dakota, Tennessee, Texas, Virginia, West Virginia, and Wisconsin.  The bulk of today’s funding will go to nonprofit rural electric cooperatives to deploy broadband throughout their service areas.   

“This latest announcement highlights the agency’s commitment to supporting even more opportunities to connect hundreds of thousands of Americans to high-speed, reliable broadband service while doing our due diligence to ensure the applicants can deliver to these unserved communities as promised,” said Chairwoman Jessica Rosenworcel.  “This program can do great things to help expand broadband in our country.” 

Together with three prior funding wave announcements, the Commission has now announced over $1.7 billion in funding to winning bidders for new deployments.  In this funding wave, 50 broadband providers will bring broadband service to over 400,000 locations in 26 states.

Biden signs Secure Equipment Act

President Biden signed into law the “Secure Equipment Act of 2021,” which requires the Federal Communications Commission to adopt rules clarifying that it will no longer review or approve any authorization application for equipment that poses an unacceptable risk to national security.

The bill would prevent further integration and sales of Huawei, ZTE, Hytera, Hikvision, and Dahua – all Chinese state-backed or directed firms – in the U.S. regardless of whether federal funds are involved.

In 2020, the FCC adopted new rules to require U.S. telecommunications carriers to rip and replace equipment provided by Huawei, ZTE, and other covered companies that pose a risk to U.S. national security. 

In October, when the bill passed the House of Representatives, the following statements were released.

Rep. Anna Eshoo (D, California): "I’ve fought for over a decade to address vulnerabilities in our telecommunications infrastructure that directly impact our national security. Equipment made by Huawei and ZTE, companies linked to the Chinese government, increase the vulnerabilities of our telecommunication systems and put the U.S. at risk. I'm so pleased that the House passed bipartisan, bicameral legislation that Rep. Scalise and I co-authored to prohibit the FCC from issuing licenses for any telecommunications equipment made by Huawei or ZTE.”

Rep. Steve Scalise (R-La.): "It was good to see both parties come together to strengthen America’s security against Chinese cyber attacks when the House overwhelmingly passed the Secure Equipment Act that I wrote with Rep. Anna Eshoo. The Secure Equipment Act, H.R. 3919, will prevent China from infiltrating America’s telecommunications networks and threatening the safety and national security of the American people when sending data across the internet. By prohibiting the FCC from issuing any equipment licenses to companies identified as a threat to our national security, this bill prevents compromised Chinese equipment from threatening America’s networks. The Secure Equipment Act sends a strong signal to the Chinese Communist Party that America is committed to securing our networks and protecting the privacy and safety of our citizens."

KIOXIA ships native Ethernet Flash-Based SSDs

KIOXIA America is now shipping its new EM6 Series Enterprise NVMe-oF™ solid state drives (SSDs) for Ethernet Bunch of Flash (EBOF) systems. 

The drives use the Marvell 88SN2400 NVMe-oF SSD converter controller that converts an NVMe SSD into a dual-ported 25Gb NVMe-oF SSD

KIOXIA says the platform exposes the entire SSD bandwidth to the network.

EM6 Series Key Features

  • Single or dual 25Gb Ethernet and RoCEv2 network connectivity
  • NVMe-oF 1.1 and NVMe 1.4 specification compliant
  • 2.5-inch 15mm Z-height form factor
  • 1 DWPD endurance with 3,840 GB, 7,680 GB capacity options