Monday, November 23, 2015

Blueprint: OpenFlow and SDN in Action

by Calvin Chai

There’s a lot of talk about OpenFlow as an evolving architecture for driving software-defined networking (SDN), but so far much of the action has been around development rather than deployments, and most of the deployment activity has been in the data center. In this article, we’ll look at some real-world examples of OpenFlow and SDN in action in the larger network.

Disaggregating the Network

Implementing open networking solutions is a matter of disaggregating the network. Historically, networking has been done through a series of vertically integrated systems. Major networking equipment vendors sell tightly integrated switches, routers, and line cards, and provisioning the network means manually configuring each network element.

In OpenFlow-driven SDN, the model is to disaggregate, to use standardized building blocks to create solutions. End users and subscribers want to change applications, raise or lower bandwidth limits, or choose the speed of their connections, and manually configured networks simply can’t respond quickly enough. OpenFlow-based SDN delivers policy-driven networks where network settings are reconfigured on the fly via the OpenFlow controller, and OpenFlow is the common control protocol between all of the disaggregated parts (network operating system, or NOS, controller, and applications).

Meter, match, and act are the three steps SDN undertakes to execute tasks in a policy- driven network. SDN enables the metering of traffic conditions, application and user behavior to match those conditions against a set of pre-defined criteria and then to act on the match according to a policy. Part of a policy framework is to pre-set conditions that are metered against. So, for example, if a customer uses a home interface to ratchet up bandwidth, the OpenFlow controller receives this input and automatically delivers more bandwidth to that subscriber. If a network must scale to support more users, the OpenFlow controller can add routes, VLANs, and flows through the network switches.

In a way, disaggregating the network is tearing it down in order to build it up again in a more useful, flexible, and dynamic way. Now, let’s look at OpenFlow and SDN applications in action.

A Subscriber Control Panel

An Asia-Pacific service provider offers its subscribers a choice of price, performance (bandwidth, latency, and throughput), and security settings. To scale this offering, the provider needed to automate the process so that network throughput, for example, could be changed remotely and quickly through a subscriber action.

Specifically, the service provider wanted to enable:

  • Liquid bandwidth – allowing subscribers to dynamically change the speed of their connections
  • Metered access – implementing an AWS-like, usage-based billing model
  • QoS on Demand – enabling the network to dynamically select the QoS level based on the usage model or application.

SDN enables these services because traffic engineering is based on policies derived from inputs to the customer portal. A graphical slider like the one in Figure 1 drives requests to the OpenFlow controller. Moreover, SDN enables seamless public and private cloud-bursting and migration, turning up more capacity automatically as needed.

Figure 1: A self-service control panel allows customers to dial up bandwidth from home or office.

The critical customer requirements for this use case were a combination of OpenFlow and legacy protocol support. Specifically the customer wanted integration between MPLS and hardware-based OpenFlow. The rationale followed what Google published in January 2011. Specifically, OpenFlow reprograms MPLS labels as a result of a policy, triggering the need to change a path for a giving traffic flow. The policy would be governed by the slider input from the control panel. From a traffic engineering perspective, paths would be chosen to ensure the path selected delivered the expected SLA.

The customer had also tracked OpenDaylight (ODL) development and felt that it was the best controller for their needs. Pica8 was the best choice for this SDN project since the Pica8 NOS, PicOS, supports OpenFlow 1.4, is integrated with ODL, and has support for MPLS.

Turning Up Bandwidth to Rural Subscribers

Renewed popularity of fiber to the premises (FTTP) for homes and small businesses is enabling municipalities and regional governments to reduce the “digital divide” and provide high-speed content and services to rural areas. Customers want to apply the “App Store” model at home for selecting content and bandwidth services, but bandwidth requirements and capabilities vary widely across the region.

Rather than manually provisioning VPNs to customer premise equipment (CPE), this regional service provider used OpenFlow to set up virtual CPE (vCPE) that can dynamically adjust bandwidth and services under user control. Two other services users desired were on-demand disaster recovery (disaster recovery as a service, or DRaaS) and hybrid cloud services for small businesses.

In this case, the vCPE can offer advanced firewall services in addition to dynamic VPNs offering DRaaS. Scalable liquid bandwidth is guaranteed as part of an SDN access network.

For this SDN deployment, the customer was keen to offer “smart city” services by leveraging existing infrastructure to minimize CapEx of new SDN services. At a fundamental level, service redundancy was desired to ensure Telco-like resiliency, thus redundant hardware was used throughout the topology. Hardware-based OpenFlow, driven again by ODL, was selected based on the customer’s involvement with this community.

Lastly, PicOS was chosen as it supports a rich set of Layer-2 and routing features including OSPF and BGP. This support ensured integration of the service edge based on white box switches with upstream Cisco and Juniper aggregation and core devices. The integrator for these smart cities additionally leveraged virtual CPEs (vCPEs) on hardened services to deliver high touch services on top of the liquid bandwidth services.

Maintaining Competitive Video Service

A North American satellite broadcast media provider faced increasing competition from over-the-top (OTT) providers and other cable and satellite broadcasters, and wanted to improve the speed, ease, and quality with which its services could be delivered. Customers were demanding an app-store, point-and-click content selection model, but a manually provisioned network couldn’t deliver this functionality. Self-provisioned networks would meet the demand, and they would also reduce the need for truck rolls and significantly lessen churn caused by long waits for new services or upgrades.

The solution was to implement self-service provisioning with point-and-click content selection via OpenFlow. The tight integration of OpenFlow and the current multicast environment drives rapid ROI and ensures smooth integration.

In this last deployment scenario, the customer was particularly interested in leveraging hardware-based OpenFlow in conjunction with the ONOS controller from ON.Labs, now governed by the Linux Foundation. As with the other two examples, integrating the new SDN functionality with legacy protocols was key. Here, since this is a media / video streaming application, multicast support was mandatory and PicOS has a suite of multicast support protocols, including IGMPv3, PIM-SM and PIM-SSM.

Lastly, integrating the new SDN functionality into existing OSS / BSS frameworks was accomplished through standard APIs.

OpenFlow is Driving Real Applications Today

As we have seen, OpenFlow-based SDN applications are increasingly driving customer satisfaction and service provider revenues in the market today. Networks are becoming more flexible in an era of on-demand computing, bandwidth, and services; and megatrends like mobile usage, Big Data, and the Internet of Things will continue to drive a need for network scale and agility. By breaking monolithic networks apart and building solutions with OpenFlow and SDN, network operators can be ready for the future.

About the Author

Calvin Chai is the Head of Product Marketing for Pica8 Inc. Prior to Pica8, Calvin has held numerous leadership roles in product, technical, solution, and corporate marketing.  Most recently, he led the cloud product marketing team at Juniper Networks where he was responsible for the marketing strategy focused on the cloud and data center markets.  Prior to Juniper, Calvin spent ten years at Cisco focused on various technology initiatives including integrated security, policy, and identity management solutions.

Calvin holds a BS degree in Computer Science and Engineering from the University of California at Berkeley.

Zayo to Acquire Allstream for its Canadian Network

Zayo Group Holdings agreed to acquire Allstream,a wholly owned subsidiary of MTS, for CAD $465 million, representing a pre-synergized adjusted EBITDA multiple of less than five times. Allstream has approximately CAD $600M revenue and Adjusted EBITDA (excluding restructuring charges) of approximately CAD $100M.

Allstream is a Canadian leader in IP communications and the only national provider that works exclusively with business customers.  Zayo said the acquisition will add substantial fiber and colocation assets across Canada to its own core network Communications Infrastructure business. Allstream has over 9,000 route kilometers of metro fiber network concentrated in Canada’s top five metropolitan markets, (Toronto, Montreal, Vancouver, Ottawa, and Calgary) that connect to approximately 3,300 on-net buildings. In addition, Allstream has an approximate 20,000 route kilometer long-haul fiber network connecting all major Canadian markets and 10 U.S. network access points. In addition, Allstream operates colocation space in Toronto, Montreal, and Vancouver.

“Within today’s Allstream is a robust collection of fiber networks, which are enormously valuable to both Allstream and Zayo customers,” explained Dan Caruso, chairman and CEO of Zayo. “We will unleash the full potential of these assets by combining them with Zayo’s network and focus on providing high-quality and low-cost bandwidth to help fuel the growth of Canada’s economy.”

Zayo estimates that approximately half of Allstream’s revenue is a direct fit with Zayo’s existing core business. Zayo’s investment thesis is to separate this business from other parts of Allstream, and integrate it into Zayo, using the same approach as is currently in place for Zayo UK and Zayo France. This plan includes retaining a strong Canadian brand and presence. The segmentation of the Communication Infrastructure portion of Allstream’s business (“Zayo Canada”) and follow-on reporting into Zayo’s core business segments (Dark Fiber Solutions, Colocation & Cloud Infrastructure, and Network Connectivity) will take multiple quarters to complete.

“As we stand up Zayo Canada, we are targeting CAD $300M of revenue, a >40 percent EBITDA margin, and a high single digit growth rate,” said Karl Maier, president of Zayo International. “If we achieve this outcome and apply an EBITDA multiple similar to Zayo, the value of Zayo Canada will be substantial.”

The other half of Allstream’s business will be organized into two additional segments: Voice and Universal Communications (approximately one third of Allstream’s revenue), and Small Business (primarily enterprise voice). Each of these will be separated into standalone business units in parallel with the formation of Zayo Canada.

MTS is Manitoba’s leading communications company and is wholly owned by Manitoba Telecom Services Inc. (TSX: MBT).

AWS Extends AWS Direct Connect in Dallas and London

Equinix announced availability of Amazon Web Services (AWS) Direct Connect in its Dallas and London International Business Exchange (IBX) data centers (DA2 and LD5).

AWS Direct Connect availability in London marks the second Equinix location in Europe to offer the service.  This deployment allows Equinix to improve upon previous connectivity options by offering “native” AWS Direct Connect service to customers into the full portfolio of AWS Cloud offerings.  Previously, connectivity in London was available via tether from Equinix in London into AWS’s Dublin facilities.  In October 2014, Equinix announced availability of AWS Direct Connect in the company’s Frankfurt data center.

Equinix’s Dallas IBX, DA2, is the fourth data center in North America to offer AWS Direct Connect, joining Equinix’s facilities in Seattle, Silicon Valley and Washington, D.C. Equinix now offers AWS Direct Connect in ten global locations; Dallas, Frankfurt, London, Osaka, Seattle, Silicon Valley, Singapore, Sydney, Tokyo and Washington, D.C./Northern Virginia. Equinix customers in these metros experience lower network costs into and out of AWS and take advantage of reduced AWS Direct Connect data transfer rates.

China Telecom Expands Rollout of ALU's 7950 XRS IP Core Router

China Telecom is deploying Alcatel-Lucent's IP routing technology, including the 7950 XRS IP Core Router and 7750 Service Router, to expand its 4G LTE network.  Financial terms were not disclosed.

Alcatel-Lucent said its 7950 XRS, which is already in use in Jiangsu province and Shanghai, will manage the increasing traffic in the urban data centers of Beijing  as the financial hub of Guangzhou. Alcatel-Lucent’s 7750 Service Router will be deployed across seven major provinces.

All deployments will be completed in preparation for service in China Telecom’s network in December.

Alcatel-Lucent is also providing 15 percent of China Telecom’s new-build edge routing capabilities under this contract, including the deployment of its 7750 SR portfolio across seven major provinces to aggregate traffic at the metro network edge.

Alcatel-Lucent now providing 25 percent of China Telecom’s new build core routing network capabilities under this contract, including support for all data center interconnections in Beijing and Guangzhou.]

Spirent Announces Customer and Network Analytics Solution

Spirent Communications unveiled an operator analytics solution that focuses on customer experience assurance and troubleshooting, with support for 2/3/4G technologies including VoLTE and the Internet of Things (IoT).

Spirent's InTouch Customer and Network Analytics (CNA) enables mobile operators’ engineering, customer care, and marketing groups to proactively identify and resolve wireless customer experience issues spanning LTE/4G networks and services like VoLTE. The solution is an evolution of the company's field-tested InTouch platform, which has been deployed in networks exceeding 100 million subscribers.

“Our solution is unique in that it allows operators to build and leverage quality of experience (QoE) scores using data mining techniques, which can prevent churn,” said Frank Galuppo, general manager of Spirent’s CEM business unit. “This allows them to rapidly identify and resolve issues before customers complain or leave.”

Several operators are already deploying beta versions of InTouch CNA, especially for new services like VoLTE and IoT.

Broadband Forum Completes FTTdp Work

The Broadband Forum completed work on the management of Fiber to the Distribution Point (FTTdp), which aids in the management of fiber-fed nodes in the periphery of the access network.

At the Broadband Forum’s quarterly meeting in Mexico, it was agreed that the FTTdp YANG management model was ready for release exclusively to members for testing in network equipment. This is the Forum’s first software project written in the YANG modelling language.

“This is great news for our members whether they be service providers, vendors or test houses,” said Broadband Forum CEO Robin Mersh. “By adopting YANG modelling we are in the process of moving the management of FTTdp from the drawing board into the network, helping to drive open interoperability between different devices. This means service providers can offer competitive ultrafast services such as”

Regarding, the Forum approved the University of New Hampshire Interoperability Labs (UNH-IOL) as a test facility for the latest plug-fest – paving the way to the industry’s first certifications.

“The success of the Forum in accelerating broadband innovation would not be possible if it weren’t for the expertise and energy of its members,” said Kevin Foster, Chairman of the Broadband Forum. “When the Broadband Forum started there were less than a million broadband connections and now there are more than 750 million. This is an extraordinary achievement and one that would not be possible without the hard work and commitment of our members, like Les, on our various projects.”

Infinera and EBlink Partner on Mobile Fronthaul

Infinera and EBlink, a supplier of of wireless fronthaul technology, announced a partnership to deliver both fiber and wireless based solutions and to provide an end-to-end fronthaul architecture.

The joint solution has already been demonstrated in a live field trial with Orange in western France that interconnected baseband units (BBU) and remote radio heads (RRH) from several of Orange's Radio Access Network (RAN) vendors. In the live field trial, the combined Infinera TM-Series and EBlink FrontLink solution demonstrated on Orange's network how it can deliver a unique end-to-end performance combining fiber and wireless for mobile fronthaul.

Where fiber exists, the Infinera TM-Series delivers bandwidth using WDM. Where there is no fiber, EBlink's wireless fronthaul is complementary to the WDM fronthaul solution.

"Mobile fronthaul is a challenging environment with tough requirements on latency and synchronisation for optical solutions," said Sten Nordell, CTO of Infinera's Metro Business Group. "As one of the few suppliers capable of achieving these requirements, we are very pleased to partner with EBlink and Orange to demonstrate that the combination of fiber and wireless is the right alchemy for mobile fronthaul."

"This partnership with Infinera reaffirms the relevance of our wireless fronthaul technologies and how complementary fiber and wireless can be," said Eric Sèle, deputy CEO of EBlink. "The deployment of the Infinera and EBlink solutions on Orange's network underscores the concept that wireless fronthaul picks up where fiber leaves off."

Brocade Posts Q4 Revenue of $589 Million, up 4% YoY

Brocade reported fourth quarter revenue of $589 million, an increase of 4% year-over-year and 7% quarter-over-quarter. Revenue for fiscal year 2015 was $2,263 million, up 2% year-over-year. The resulting GAAP diluted earnings per share (EPS) was $0.20 for the fourth quarter and $0.79 for fiscal year 2015, up 6% and up 48% year-over-year, respectively.

"Fiscal 2015 was a productive year in which we achieved many significant milestones," said Lloyd Carney, CEO of Brocade. "We delivered annual revenue growth in fiscal 2015, with a year-over-year revenue increase in each fiscal quarter. We grew our non-GAAP EPS by 12% for the fiscal year, delivering more than a dollar per share for the first time. We continued to expand our portfolio of software and hardware products through both technology innovation and strategic acquisitions. Looking forward, these investments create new opportunities for us to continue to grow revenue and EPS in 2016 and beyond."

SAN product revenue for Q4 was $325 million, flat year-over-year and up 5% quarter-over-quarter. The Q4 year-over-year product revenue performance reflects a 14% increase in director sales and a 1% increase in embedded switch sales, offset by a 12% decrease in switch sales. For fiscal year 2015, SAN product revenue was $1,301 million, down 2% year-over-year, primarily due to lower switch and embedded switch sales, partially offset by higher director sales.

IP Networking product revenue for Q4 was $170 million, up 12% year-over-year and 10% quarter-over-quarter. The Q4 year-over-year increase was primarily driven by a 28% increase in Ethernet switch sales and improved software sales, partially offset by a 20% decline in router revenue. For fiscal year 2015, IP Networking product revenue was $601 million, up 14% year-over-year due to stronger switch, router, and software sales.