Tuesday, June 20, 2017

Cisco's Intent-based Networking Leverages Machine Learning

Cisco introducted its vision for "Intent-based Networking", a paradigm that it says will form be the foundation for enterprise infrastrcuture for the next 30 years.  Intent-based networking will leverage machine learning in a new generation of ASIC-powered swithches to derive insight from network traffic, even if it is encrypted. These insights would be used to dynamically adjust network policies to simplify management and mitigate cyber threats.

“By building a more intuitive network, we are creating an intelligent platform with unmatched security for today and for the future that propels businesses forward and creates new opportunities for people and organizations everywhere,” said Chuck Robbins, chief executive officer for Cisco.

Intent-based networking includes:

  • DNA Center - a centralized management dashboard with an intent-based approach for full visibility and context across the entire network, DNA Center allows IT to centralize management of all network functions. 
  • Software-Defined Access (SD-Access) - uses automated policy enforcement and network segmentation over a single network fabric. Cisco said that its initial analysis with field trial customers and internal testing have shown a reduction in network provisioning time by 67%, improved issue resolution by 80%, reduced security breach impact by 48%, and opex savings of 61%.
  • Network Data Platform and Assurance - efficiently categorizes and correlates the vast amount of data running on the network and uses machine learning to turn it into predictive analytics, business intelligence and actionable insights delivered through the DNA Center Assurance service.
  • Encrypted Traffic Analytics - uses Cisco’s Talos cyber intelligence and machine learning to analyze metadata traffic patterns. The network can identify the fingerprints of known threats even in encrypted traffic, without decrypting it and impacting data privacy. Cisco claims its can detect threats in encrypted traffic with up to 99% accuracy, with less than 0.01% false positives. 
  • Catalyst 9000 Switching Portfolio - a new family of switches built from the ground up for the new realities of the digital era, centered on the demands of mobility, cloud, IoT and security. The Cisco Catalyst 9000 features innovations at the hardware (ASIC) and software (IOS XE) layers.
  • Software Subscription - DNA software capabilities are now offered by subscription either via pre-bundled Cisco ONE software suites or a-la-carte components. Available across the entire enterprise networking portfolio, Cisco ONE software provides businesses with access to ongoing innovation, budget predictability, and a more agile way to consume the technology.
  • DNA Services - a new portfolio of services, including advisory, implementation, optimization and technical services. Cisco channel partners can also resell these services and build networking practices that incorporate software, security, automation and analytics for their customers.
  • Developer Center - resources to help developers and IT professionals create network-powered applications and integrate them within their IT systems and workflows. This includes new learning tracks, sandboxes, and developer support resources for using APIs and building skills.

Cisco said these technologies are already being tested by 75 global enterprises and organizations, including DB Systel GmbH, Jade University of Applied Sciences, NASA, Royal Caribbean Cruises Ltd., Scentsy, UZ Leuven and Wipro.


The Evolution of VNFs within the SD-WAN Ecosystem

As the WAN quickly solidifies its role as the performance bottleneck for cloud services of all kinds, the SD-WAN market will continue to grow and evolve. This evolution will happen in lock step with the move to software-defined everything in data centers for both the enterprise and the service provider, with a focus on Virtual Network Functions (VNFs) and how they could be used to create specialized services based on custom WANs on demand. Although SD-WANs provide multiple benefits in terms of cost, ease-of-management, improved security, and improved telemetry, application performance and reliability remain paramount as the primary goals for the vast majority of SD-WAN deployments. When this is taken into consideration, the role of VNFs in extending and improving application performance becomes clear. Just as importantly, growing use of VNFs within SD-WANs extends an organization’s software-defined architecture throughout the broader network and sets the stage for the insertion of even more intelligence down the road.

What exactly do we mean by the term VNF? 

Before we get started, let’s define what we mean by VNF, since similar to SD-WAN, this term can be used to describe multiple things. For some VNFs are primarily a means of replicating legacy capabilities on a local appliance (physical or virtual) by means of software defined architectures, such as firewall, DHCP, DNS etc. However, restricting one’s scope to legacy services alone limits the potential high-value benefits that can be realized from a software-defined approach for more advanced features. Our definition of a VNF therefore is a superset of localized VNF and is really about the creation of a software-defined functions of more advanced capabilities, such as application aware VPNs, flow-based load balancing, self-healing overlay tunnels etc. What’s more, many advanced SD-WAN vendors provide their customers with the ability to customize these VNF applications to apply exclusively to their own WAN and/or their specific network requirements to enable unique WAN services.

What do we need VNFs for? 

SD-WAN’s enormous growth this year, as well as its predicted continued growth in the years to come follows the footsteps of the paradigm shift data centers are currently undergoing. That is, from a manually configured set of servers and storage appliances, to a software-defined architecture, where the servers and storage appliances (virtual or physical) can be managed and operated via a software-defined architecture. This means less manual errors, lower cost and more efficient way to operate the data center.

As an industry, as we implement some of the data-center approaches to the WAN (Wide Area Networks), one must note that there is a big difference between datacenter networks and WAN networks. Namely, datacenter LANs (Local Area Networks) have ample capacity and bandwidth and unless they are misconfigured, are never the bottleneck for performance. However, with WANs, whether done in-house by the enterprise or delivered as a service by a telecom or other MSP, the branch offices are connected to the Internet through WAN connections (MPLS, DSL, Cable, Fiber, T1, 3G/4G/LTE, etc.). As a result, the choking point of the performance is almost always the WAN. This is why SD-WANs became so popular so quickly, in that this provides immediate relief for this issue.

However, as WANs continue to grow in complexity, with enterprises operating multiple clouds and/or cloud models simultaneously, there is a growing need to add automation and programmability into the software-defined WAN in order to ensure performance and reliability. Therefore VNFs that can address this WAN performance bottleneck have the opportunity to transform how enterprises connect to their private, public and hybrid clouds. VNFs that extend beyond a single location, but can cover WAN networks, will have the ability to add programmability to the WAN. In a way, the “software defined” nature of the data center will be stretched out all the way to the branch office, including the WAN connectivity between them.

Defining SD-WAN VNFs

So what does a VNF that is programmable and addresses the WAN bottlenecks look like? These VNFs are overlay tunnels that can perform certain flow logic and therefore can work around network problems on a packet-by-packet basis per flow. These VNFs are so smart, they have the problem diagnosis, problem alerting and most importantly, resolution of the problem all baked into the VNF. In other words, unlike the days without SD-WAN where an IT manager would have an urgent support ticket whenever a network problem occurs. With VNF-based SD-WANs, the networks are becoming smart enough to solve the problem proactively, in most cases, before even it effects the applications, services and the user experience.

This increase in specific VNFs for the SD-WAN will start with the most immediate need, which is often latency and jitter sensitive applications such as voice, video, UC and other chatty applications. Even now, VNFs are being used to solve these issues. For example, a CIO can have a VNF that dynamically and automatically steers VOIP/SIP traffic around network problems caused by high latency, jitter and packet loss, and in parallel have another VNF to support cross-traffic and latency optimization for “chatty” applications.

In another example, a VNF can be built in minutes designed to steer non-real-time traffic away from a costly WAN link and apply header compression for real-time traffic only in situations where packet loss or latency crosses a specific threshold during certain times of the day, all the while updating syslog with telemetry data. With this level of flexibility and advanced capabilities, VNFs are poised to become the go-to solutions for issues related to the WAN.

A VNF load balancer is another such overlay that has the ability to load balance the traffic over the WAN links. Since the VNF load balancer is in essence a software code that can be deployed onto an SD-WAN appliance, it has the power of taking advantage of various types of intelligence and adaptability to optimize the WAN performance. VNF load balancers should also work with standard routing so that you can inject it in your network, say between the WAN modems and your firewall/router seamlessly.

Clearly, VNFs are part and parcel of SD-WAN next wave of evolution, bringing intelligence and agility to the enterprise WAN. As 2017 ramps up, we’ll see more and more innovation on this front, fully extending software-defined architecture from the data center throughout the network.

About the author

Dr. Cahit Jay Akin is the CEO and co-founder of Mushroom Networks, a long-time supplier of SD-WAN infrastructure for enterprises and service providers. Prior to Mushroom Networks, Dr. Akin spent many years as a successful venture capitalist. Dr. Akin received his Ph.D. and M.S.E. degree in Electrical Engineering and M.S. in Mathematics from the University of Michigan at Ann Arbor. He holds a B.S. degree in Electrical Engineering from Bilkent University, Turkey. Dr. Akin has worked on technical and research aspects of communications for over 15 years including authoring several patents and many publications. Dr. Akin was a nominee for the Most Admired CEO award by San Diego Business Journal. 

Huawei unveils OXC+OTN all-optical switch

Huawei has released its mesh backbone network solution, featuring all-optical switching, full mesh and one-hop service transmission and designed to enable the delivery of ultra-high bandwidth at the Next Generation Optical Networking (NGON) 2017 event.

The new Huawei solution features what is claimed to be the first commercial optical cross-connect and optical transport network (OXC+OTN) cluster devices and is intended to enable data centre (DC)-centric backbone networks and to provide wavelength/sub-wavelength connections between DCs for one-hop transmission, thereby reducing latency and supporting ultra-high bandwidth between any two facilities.

At NGON, Huawei demonstrated the solution providing dynamic grooming of optical wavelengths and cross-connections through OXC, as well as OTN cluster, high-integration and multi-functional service boards.

Huawei believes that inter-DC traffic will become the main type of traffic carried over transport networks, and that to address this requirement operators will need to evolve their backbone networks originally designed for traditional telecom services. New DC-centric networks will enable mesh interconnectivity between nodes and one-hop service transmission to help shorten the network path for lower latency and allow real-time exchange of data between DCs to effectively support cloud services.

Huawei's new OXC+OTN cluster mesh backbone network solution is designed to allow transport backbone networks to be re-architected for the cloud era, in particular by providing a cluster mesh backbone network that enables one-hop transmission between any two cities.

Leveraging wavelength-level switching and liquid crystal on silicon (LCOS) technology, the OXC provides from 320 up to 640 Tbit/s cross-connect capacity and supports wavelength grooming in up to 32 optical directions, while operating with power consumption at the level of hundreds of watts.

Additionally, a new optical backplane addresses the issue of complicated fibre connections within traditional ROADM. The optical backplane is designed to significantly simplify fibre connectivity, as well as reducing loss associated with connection and improving system reliability.

The OTN component of the solution serves to address access and grooming requirements of small-granularity services at the sub-wavelength level. The cluster technology supports non-blocking cross-connections between OTN subracks and provides resource pools for transmission channels, thereby enabling continuous expansion of channel resources between DCs.

Arista enhances R-Series universal spine and leaf platforms

Arista Networks has introduced new software and hardware for the Arista R-Series platforms, designed to accelerate cloud transformation initiatives for customers deploying routing technology.

Based on merchant silicon designed to deliver twice the density with half the power usage of custom routing silicon, Arista's latest R-Series platforms provide over 150 Tbit/s of capacity for switching and routing applications with cloud-driven Arista EOS software technologies including Arista FlexRoute and AlgoMatch.

The company noted that following the introduction of the Arista 7280R Universal Leaf and 7500R Universal Spine platforms last year, cloud transformation is underway for multiple routing use cases by cloud providers, content delivery networks (CDN) and ISPs. Arista is now focusing on the next phase of this transition via the introduction of 7500R2 line cards for the 7500R systems and 7280R2 fixed platforms, designed to enable increased scale, telemetry, security and routing.

Noting that cloud network designs require both scaling up and scaling out, Arista's FlexRoute enhancements on the 7500R2 line cards and 7280R2 fixed platforms provide support for over 2 million routes in hardware, equivalent to more than three copies of the Internet routing table. As the Internet routing table continues to grow and native IPv6 networks increase, customers can adopt a routing platform able to handle the Internet table scale into the future.

The latest member of the 7500R Series, the 7516, offers greater port scale, with up to 576 ports of line rate 100 Gigabit Ethernet, designed to provide the basis for scale-out network designs, including spine capacity of up to 15 Petabit/s.

Additionally, AlgoMatch on the 7500R2 line cards and 7280R fixed platforms introduces sFlow for enhanced traffic analysis in routing use cases. Designed to allow scaling of telemetry at 100 Gbit/s interface speeds, the solution is claimed to make the R-Series the highest visibility solution for dense 100 Gbit/s applications.
Backwards compatible with the 7500R modular platform, the new capabilities offered with the 7500R2 line cards and 7280R2 systems include:

1.         16-slot 7500R Series providing higher density spine switches with up to 15 Pbit/s and 576 ports of 100 Gbit/s in a single system, with the flexibility to support 1 to 100 Gbit/s with future support for 400 Gbit/s.

2          .Expanded high capacity 7280R2 with up to 60 ports of 100 Gbit/s, IEEE 802.3by 25 Gigabit Ethernet and over 2 million routes in hardware.

3.         sFlow export on both fixed and modular systems.

4.         New data centre interconnect solutions with secure 802.1AE MACsec encryption on 36-port 100 Gbit/s line cards and coherent 200 Gbit/s DWDM line cards with reach of over 5,000 km.

Arista's 7500R2/7280R2 Series are available immediately; the 7516 is currently in customer trials, with general availability due later in 2017.

Ciena to upgrade Bouygues Telecom backhaul, aggregation, SuperCore

Ciena announced that Bouygues Telecom of France, serving over 3 million fixed broadband and around 13 million mobile subscribers, has selected its 6500 Packet-Optical Platform to help improve the efficiency, scalability and reliability of its current network backhaul and aggregation infrastructure, as well as to support the creation of its new SuperCore network.
Deployment of the Ciena platform is designed to increase capacity to up to 400 Gbit/s to support increasing capacity demands and new high-bandwidth services, as well as allow Bouygues Telecom to continue to expand its mobile and ISP offerings and establish the foundations for a future 5G network.

This network upgrade will also support Bouygues Telecom's efforts to offer customers improved quality of service on its 4G network and also enable the company to attain its goals for the fixed network, including through the expansion of FTTH services.

Earlier this year, Bouygues Telecom and Ericsson announced a demonstration of 5G technology at the operator's Technopôle complex in Meudon-la-Forêt, France. The demonstration featured prototype mobile devices simultaneously connected to an Ericsson radio base station equipped with active 5G antennas.

During the tests, a team from Bouygues Telecom and Ericsson performed simultaneous data transfers and 4K video streaming to the two devices with aggregate throughput of over 25,2 Gbit/s via the 5G antenna system, with a latency of around 3 ms. The demonstration specifically involved adaptive beamforming technology, combined with massive MIMO at centimetre-wave frequencies.

Finisar reports Q4 revenue of $357.53m, up 12.1% yr/yr, net income of $130.24m

Finisar, a major supplier of subsystems and components for fibre optic communications, announced financial results for its fourth quarter and full fiscal year 2017, ended April 30, 2017, as follows:

1.  Revenue for the fourth quarter of 2017 of $357.53 million, down 6.1% compared with $380.5 million in the third quarter of fiscal 2017 and up 12.1% from $318.79 million in the fourth quarter of 2016.

2.   Gross profit for the fourth quarter of $125.16 million, down 8.3% compared with $136.64 million in the third quarter of fiscal 2017 and up 38.4% from $90.44 million in the fourth quarter of 2016.

3.  R&D expenditure for the fourth quarter of $58.97 million, up 8.0% compared with $54.59 million in the third quarter of fiscal 2017 and up 17.5% from $50.17 million in the fourth quarter of 2016.

4.  SG&A expenditure for the fourth quarter of $24.64 million, down 6.4% compared with $26.33 million in the third quarter of fiscal 2017 and down 3.3% from $25.47 million in the fourth quarter of 2016.

5.  Total operating expenditure for the fourth quarter of $84.32 million, up 3.2% compared with $81.73 million in the third quarter of fiscal 2017 and up 10.5% from $76.31 million in the fourth quarter of 2016.

6.  On a GAAP basis, net income for the fourth quarter of $130.24 million, compared with net income of $46.39 million in the third quarter of fiscal 2017 and net income of $13.07 million in the fourth quarter of 2016.

On a non-GAAP basis, net income for the fourth quarter of $57.51 million, compared with net income of $67.20 million in the third quarter of fiscal 2017 and net income of $31.82 million in the fourth quarter of 2016.

7.  Cash and cash equivalents as of April 30, 2017 of $260.23 million, versus $240.59 million as at January 29, 2017 and $299.22 million as at May 1, 2016.

Additional results and notes

For the full year 2017, Finisar reported revenue of $1,449.3 million, versus $1,263.2 million a year earlier, with net income of $249.35 million, compared with net income of $35.19 million in fiscal year 2016.

Finisar noted that in the fourth quarter, sales of datacom products increased by $2.8 million (1.1%) compared to the third quarter, with sales of 100 Gbit/s QSFP28 transceivers for datacom applications up approximately 30% sequentially, but offset by lower demand for other datacom products, mainly 10 Gbit/s and below shortwave transceivers.

Sales of telecom products in the fourth quarter decreased by $20.2 million (18.2%) compared to the third quarter, primarily due to lower revenue from Chinese OEM customers and 3 months of annual telecom price erosion.


For the first quarter of fiscal 2018, Finisar currently expects revenue in the range $330 to $350 million, representing a sequential decline of 4.9% at the midpoint.

Australia's nbn begins roll-out of FTTC technology, aims to reach 1m premises by '20

nbn of Australia announced it has commenced the initial stages of deploying its advanced fibre to the curb/kerb (FTTC) access technology, which is expected to serve more than one million homes and businesses nationwide by 2020 as part of its national broadband network (nbn).

Designed to enable faster broadband speeds through extending fibre closer to customers, the FTTC technology will provide access to services over the nbn network by delivering fibre to the driveway of the home.

nbn stated that is also working with retailer customers that will offer the new service to end users to develop a FTTC product, which is scheduled to be available to consumers and businesses by mid-2018.

The nbn network is currently available to almost half of Australia's population, with plans to reach three quarters of the country by the middle of next year and to complete the roll-out by 2020.

Key regions where nbn plans to commence construction of the FTTC network during the second half of 2017 include: New South Wales – 96,100 premises in Sydney/greater Sydney; Victoria – 89,000 premises in Melbourne/greater Melbourne; Queensland – 44,200 premises in Brisbane/greater Brisbane; ACT – 4,600 premises in greater Canberra; South Australia – 22,800 premises in Adelaide/greater Adelaide; Western Australia – 9,000 premises in Perth/greater Perth.

  • nbn recently announced that it had reached a total of 5 million home and business premises with its wholesale broadband network, and that there were a total of 2.2 million customers connected to the network.

Corning launches multiuse platform for deployment of fibre-deep access networks

Corning announced its new multiuse platform, which is claimed to be the first solution to provide a combination of multi-fibre and single-fibre connection points to help carriers, operators and municipalities deploy fibre-deep access networks.

Capable of supporting a range of network architectures in the same fibre backbone, Corning's multiuse platform is designed to allow service providers to expand and enhance LTE connectivity and prepare for the transition to 5G in the future, and encompasses key elements.

New local convergence cabinet

The local convergence cabinet is significantly smaller and lighter than standard industry offerings and features a range of splitter types and pass-through fibres to facilitate deployments across any network connection point, and enable a fully pre-terminated experience in the distribution network, removing the need for field work. Technicians can plug in feeder and distribution cables with factory-installed, hardened connectors, with no field splicing required from the cabinet out.

Expanded range of network access points

An expanded array of factory-installed, sealed network access points with single-fibre and multi-fibre options are designed to enable fast, plug-and-play connectivity, allowing network operators to deliver a mix of fibre counts suitable for residential, business, wireless backhaul and wholesale purposes from the same network location.

By eliminating splice points and offering an advanced plug-and-play cabinet, the Corning multiuse platform is claimed to deliver up to a 50% saving in labour costs compared with field-termination methods, along with the assurance of factory-terminated performance.

Additionally, by enhancing the capabilities of traditional pre-connectorised solutions, the new multiuse platform is claimed to reduce total installed cost by up to 20%.

LightCounting: 100G QSFP28 up 40% to >$250m in Q1

LightCounting, in its latest Quarterly Market Update Report, finds that demand for 100 Gigabit Ethernet optical connectivity from operators of mega-data centres is continuing to exceed supply, and estimates that sales of QSFP28 SR4, PSM4, CWDM4 and LR4 transceivers rose 40% in the first quarter to over $250 million.

The research firm reports that overall sales of Ethernet, SONET/SDH, CWDM/DWDM, Fibre Channel, FTTx, wireless fronthaul transceivers and active optical cables (AOCs) and embedded optical modules (EOMs) declined 5% sequentially in the first quarter, but increased 15% year on year. It forecasts that weak demand for optics in China will result in a further 3-5% decline in the sales of these products in the second quarter, with sales of tunable lasers, modulators and coherent receivers set to decline by 20% or more.

More specifically, LightCounting reports that suppliers of optical components and modules experienced a sharp fall in orders from Huawei and ZTE in March 2017, after ZTE reached a settlement with the U.S. government relating to an investigation into violations of export sanctions to Iran. It notes that the investigation, launched in 2016, could have restricted ZTE's access to optics from U.S.-based suppliers, so that the vendor may have accumulated excess inventory of such products.

While Huawei was not formally charged with such violations, the research firm believes that it may also have started to build up reserves of key components from U.S. suppliers to avoid potential disruptions. When the settlement was reached, both companies may have started to use components from their reserves and reduced purchases of new products.

LightCounting states that if this view is correct, demand for optics in China would be expected to return shortly as excess inventory is depleted. It adds that there were signs of rising demand for optical components in June, which is expected to continue in July and August.

LightCounting notes that ZTE and Huawei both reported higher sales of optical networking equipment in the first quarter and issued strong guidance for the current quarter. It also cites Chinese service providers as making ongoing investments in optical networking even as overall capex is reduced in 2017. Additionally, there are multiple new projects scheduled for the second half of the year in China.

Finally, LightCounting reports that shipments of 200 Gbit/s DWDM ports reached a new record in the first quarter, while DWDM solutions enabling 400 Gbit/s transmission on a single wavelength are an area of focus. In addition, the development of 200/400 Gigabit Ethernet transceivers is also underway.