ONOS: The Game Changing SDN Controller?


ONOS (Open Network Operating System) is the first Open SDN Controller with focus on Service Provider Use Cases and their challenges of Scalability, High Availability & Performance. The support of Users and growing ecosystem, along with promising performance results make it a strong contender among SDN Controllers.

Software Defined Networking (SDN) has seen unprecedented hype and industry activity in the last three years.

“SDN is the next big thing in networking in decades, perhaps the next big thing after networking itself”.

Dan Pitt, ONF Executive Director

Indeed SDN brings a ray of hope for stagnating Telecom industry that might not have a sustainable business model with status quo in the future. As a result, several approaches to SDN have emerged each of its pros and cons. Among the available SDN Controllers, we have both vendor proprietary and open-source controllers. Vendor Specific controllers include VMware Nicira NSX, Cisco APIC, Alcatel-Lucent Nuage VSC, Brocade Vyatta, HP VAN and Cyan Blue Planet etc., while most notable open-source controllers are Open Daylight, Open Contrail, Big Switch Project Floodlight and Ryu etc. ONOS (Open Network Operating System) is the latest addition in the list of Open Source SDN Controller Projects. ONOS has been under development by ON.Lab (Open Networking Lab) since 2012 and was Open Sourced on Dec 5th 2014 after the support of AT&T and NTT in Nov 2014.

After having so many existing open-source controllers, one might wonder the need for another and why is it important?

To answer this, we have to look at the motivation behind open source projects and analyze various projects to see how they can deliver on the original SDN objectives of lower costs (both CAPEX & OPEX), faster TTM, Innovation, separation of control and forwarding and avoidance of vendor lock-in.

Open-source projects bring benefits such as Interoperability, Flexibility & Freedom from vendor lock-in, Customizability, reduced Costs and support from a large Community. There are three types of Open Source Ecosystems: Vendor Dominated, Loosely Aligned and Foundation Managed. The most effective and impactful projects tend to be Foundation Managed, e.g. OpenStack & Apache. The reason being a strong and independent community that is led by clear customer requirements, value proposition and transparency in rules of engagement among the members.

Majority of the existing SDN open-source initiatives are vendor dominated with inherent associated risks which could diminish the value of open-source, like vendor lock-in and business-model favoring the dominating vendor. In other words, such projects could be seen quite close to vendor proprietary solutions with the exception of added customization and interoperability.

One exception is OpenDaylight, which is a Linux Foundation project, supported by the largest community of vendors and has released its second code base Helium in March 2015. However, there has been skepticism over its intentions fueled by the fear that it was dominated by heavyweights like Cisco and IBM. ONF’s Dan Pitt also expressed his concerns that OpenDaylight seemed to have slowed down the SDN market innovation and it did not involve any users which was critical to understand and meet real user needs. ONF has also been critical of Cisco’s proposed OPFlex protocol which is now expected to emerge in OpenDaylight’s upcoming “Lithium” release.

In this context, ONOS has emerged as the first Open-Source SDN project focusing exclusively on service provider use cases and led by users including operators like AT&T, NTT, SK Telecom and ONF which has the support of software-oriented companies such as Google, Facebook, Yahoo, Verizon, and Microsoft. ONOS promised to address the challenges of Scalability, High Availability & Performance, all of which are critical for Carrier-Grade Networks.

“Software-defined networking can radically reshape the wide area network, the introduction of ONOS provides another open source SDN option designed for service provider networks with the potential to deliver the performance, scale, availability and core features that we value.”

John Donovan, SEVP, AT&T Technology & Operations.

While OpenDaylight has tried to accommodate several protocols in its southbound API layer, ONOS is focusing primarily on OpenFlow for new deployments and Netconf for legacy equipment. It also has Distributed (Clustered) Core which to provide scale-out and solve the problem of Open Flow controller scalability. It has a three month release cycle (more aggressive than ODL), the code for its second release, Blackbird, can be downloaded here.

ONOS Distributed Architecture

Given the support of large operators, vendors are rapidly joining the ecosystem and current list includes: Ericsson, Ciena, Cisco, Huawei, Intel, NEC and Fujitsu. Among them, Huawei has announced full support of ONOS in their SDN solution, while Ericsson terms ONOS as the cure for performance anxiety in SDN. Recently, ONOS has released performance metrics for its Blackbird release.

The numbers related to ONOS Blackbird performance metrics are impressive and promising.

In addition to performance, ONOS has several Service Provider Use Cases including Segment Routing, NFaaS, Multi-layer SDN and SDN-IP among others. It provides a vision of future network where POP and CO’s will be built like Data Centers and Core and Metro Packet Optical networks will be built on low cost hardware with separation of control and forwarding.

ONOS Service Provider Future Network


ONOS and OpenDaylight are the two most promising Open-Source SDN Controllers on the horizon but it’s still early days for both as yet. While ODL is arguably more mature and have a larger ecosystem (even though recently Juniper and VMWare reduced their commitments), ONOS has the increasing support of Operators and Use Cases which are customized for Operators’ requirements. Given the pace of rapid developments in the industry, it might take some time for a clear winner to emerge. The key to long-term success of either will be the adoption of real-world deployments in large-scale service provider networks.

Starting with a vendor-dominated/proprietary solutions can be a great tactical choice to understand technology, build internal capability and deploy current and immediate SDN Use Cases. However, given that SDN is going to be strategic to service provider’s core business, any SDN ecosystem should be evaluated to ensure vendor neutrality, large and non-vendor dominated base of developers & integrators and long-term viability (industry support) of the ecosystem.

Disclaimer: This article is an analysis on current and emerging state of SDN controllers and is not meant to cover a comprehensive list of SDN solutions/providers. The aim is to list examples to provide a clearer understanding but not to favor any company/vendor. All logos used in the article belong to their respective trademark owners.


How NFV impacts Content Delivery Strategies of Telcos/CSPs


CSPs have an opportunity to utilize NFV to their advantage and exploit their proximity to end user by pushing the content closer, optimizing delivery costs and user-experience and also working out better content monetization models.

I have previously written about how the emergence of network cloud or more specially SDN and NFV brings a ray of hope for stagnating telecom industry and some of the technology considerations as well as organizational transformation challenges that need to be addressed for successful attainment of the potential benefits.

We know that global traffic consumption is increasing exponentially and even doubling every year in some markets. The majority of the traffic today in Internet video content, Cisco 2014 VNI predicts that video will constitute up to 78% of the global Internet traffic by 2018. Efficient delivery of this video dominated content is a challenge for current networks on both technical and commercial fronts. From technology perspective, traditional Internet was designed to deliver packet flows, which is not yet optimized for video traffic. There is no integration of application layer intelligence along the value chain as well as no end-to-end QoS mechanisms. On commercial side, content owners keep on developing higher quality content because on low delivery costs and telecom operators are faced with the dilemma of continuously expanding bandwidth without any clear ROI due to the existing interconnection and peering models.

The result is that various Content Delivery Networks (CDNs) have been deployed to improve efficiency and content quality. The CDN market had evolved rapidly over the last decade with CSPs, Internet Players/Content Providers (like Google, Yahoo, Microsoft etc.) and even new Pure-play CDN providers (like Akamai, Limelight, Level 3, CDNetworks etc.) all building their own content distribution networks. As per Cisco 2014 VNI, more than 55% of total Internet traffic will be CDNized by 2018.


However, traditional CDNs lack integration down the service provider networks or up with the Applications to ensure E2E QoS & SLA. CSPs/Service providers have to deliver two kinds on content: CSP (managed) content and non-CSP (unmanaged) content. Managed content has to be delivered On-Net (for consumers within the network reach of CSP) and Off-Net (for consumers on other CSP networks). Similarly, unmanaged content can be CDNized or Non-CDNized. As a result, there are multiple models for optimizing content delivery for CSPs:

Transparent Caching

In a bid to improve average response time for their customers and reduce the traffic on expensive international links, several operators deployed transparent Internet caching. However, its utilization is impacted by cache-ability and homogeneity of the content. With HTML5 and Internet 2.0 more and more content is dynamic which reduces caching efficiency and makes the business case for stand-alone caching quite unattractive.

CDN Collocation

This is the basic entry model of OTT and CDN players by collocating their nodes to the edge of service provider networks. This is the most commonly practiced model by Telcos where integration is simple and costs are low. This model can be with or without payments to Telcos who also benefit from improved user experience, however they lose out on majority of the revenue generated by OTT/CDN players for content delivery value chain. Simple collocation at the edge of the network doesn’t add enough value for OTT/CDN players to convince them to share revenue and Telcos still can’t ensure flawless experience for their customers.


Another model is to build own CDN to efficiently deliver the content. This approach focuses more on generating revenue and monetization of content. However, the CDN footprint and the amount of content originating from that footprint impact the business case. Several operators have gone down this route but they have to compete with global CDN players who have extensive presence all over the world and are able to offer very competitive offerings. Hence, the business case is dubious for small to medium sized operators.

CDN Resell

As opposed to building own CDN, this model enables CSPs to white-label commercial CDN offerings to sell to their Enterprise customers without the required investment on infrastructure. It provides additional revenue opportunity for Telcos as well as CDN players. Deployment is very quick and the risk to either is quite low, that is the reason the many Tier-1/2 service providers are adopting this approach. However, this model limits CSP value as a marketing arm of CDN player. Other than customized product bundles, there is not much value that CSPs offer the Enterprises and there is nothing stopping them to go directly with CDN players as this way they can pick and chose the best offer for their requirements.

CDN Interconnect/Federation

Today all CDNs are working as silos with no way for content providers to ensure E2E QoS/SLA for their content, which may traverse across multiple networks/CDNs before reaching the end customers. The concept of CDN Interconnection (CDNI) or CDN Federation is aimed to provide seamless content delivery with same QoS parameters across multiple networks. Standardization bodies like IETF and ETSI are working on use cases, requirements, architecture and protocols for CDNI. However, there are several challenges still to be addressed before a CDN federation can become a reality. On technical front issues like provisioning across multiple CDN networks, consistency in configuration, change management, capacity management, consolidated reporting, billing and NOC-to-NOC fault management exist, while on the commercial side relationship management, differing regulatory regimes and simultaneous cooperation & competition need to be addressed.

Deeper CDN Integration and Virtualized CDN

An opportunity exists for deeper integration and interworking between OTT/Commercial CDNs and Telco networks. The promise is superior customer experience, more control of content and an opportunity for new business models for content delivery. However, with the number of global CDNs reaching dozens, it will be impractical and cost prohibitive to push delivery nodes from each CDN down to network nodes at access/aggregation (PoP) level. The capacity of each dedicated physical appliance will need to be dimensioned for peak traffic, the appliances will not be able to react to unforeseen traffic needs beyond their capacity and the physical appliances increase the network complexity and OPEX.

This is where Network Functions Virtualization (NFV) comes to the rescue. It will be very easy to host multiple virtualized instances (VMs) of vCDN platforms on a standard x86 based COTS (Commercial off-the-shelf) platform. Now CSPs can have superior QoE for their customers than simple CDN Collocation while utilizing resource sharing, cost efficiency and scale-out capability for unforeseen traffic. CSPs will also have the leverage to host multiple CDN services on the same infrastructure pushing out all CDNs towards the end user and freeing up significant upstream network capacity from CDN and gain the capability to offer new wholesale business models and CDN select. It is also argued that NFV could unify CDN fragmentation.

This is due to these potential benefits that ETSI has highlighted Virtual CDN as one of the major use cases of NFV and some Tier-1 Telcos have also started pushing for vCDN VM hosting rather than physical node collocation.


Principle of different vCDN cache nodes deployment in Virtualized environment (Source: ETSI)


Optimizing delivery of ever expanding content is an absolute necessity for CSPs. Multiple models and strategies exist, starting from simple transparent caching to building their own full scale CDN. However, given the volatile nature of content delivery market with increasing number of content sources, formats, protocols and device types, a more inclusive approach might be a sweet spot for CSPs to address both managed and unmanaged content types in both on-net and off-net delivery modes. Cost efficiency, resource sharing, service agility and scale-out operations are becoming extremely important components of a successful content delivery network. CSPs have an opportunity to utilize NFV to their advantage and exploit their proximity to end user by pushing the content closer, optimizing delivery costs and user-experience and also working out better content monetization models in the process.

Strategies for Future Mobile Backhaul Networks in the era of LTE-A, Small Cells and Cloud-RAN

This article tries to summarize various technologies and media types used for mobile backhaul and explore the impact of LTE-A, Small Cells and Cloud-RAN on Mobile Backhaul/Fronthaul Strategies.

Mobile backhaul has evolved significantly over the last few decades. During initial deployment of 2G/GSM networks in 1990s, backhaul was provided using TDM (E1/T1 circuits). Introduction of 3G NodeBs was supported by ATM circuits in the beginning and then the industry settled at IP/Ethernet backhaul. The result was a hybrid or dual backhaul approach with both TDM and IP/Ethernet. With the introduction of LTE eNodeBs over the last five years or so, the backhaul traffic mix has been tilting more and more towards All IP/Ethernet.


Similarly from the choice of medium perspective, there have been three dominant media types:

  1. Microwave (offering ease of installation and lower CAPEX)
  2. Copper (utilizing existing telephony infrastructure)
  3. Fiber (providing the scale of bandwidth necessary for LTE and even LTE-A, thought as the most future proof medium).

As the penetration of different mobile access technologies has evolved, we have seen fiber replacing copper access in many areas, however microwave has maintained its share because of costs involved in laying fiber to each remote cell site. This distribution has also been impacted by geography, terrain and economical context of various countries and regions.

Below is an indicative graphic to represent the complexity of mobile backhaul that this mix of technologies and media types has induced over the last several years. It is not meant to be comprehensive from solution availability point of view but rather cover some of the more famous technologies which are being used for mobile backhaul.


Increasing requirements for LTE-Advanced

LTE advanced is a combination of enhancements over LTE including features like CA (Carrier Aggregation), CoMP (Coordinated Multipoint) and eICIC (enhanced inter-cell interference coordination). Each of them brings the benefits of increased throughput and/or coverage but introduces new challenges for backhaul.

  • Peak bandwidth requirement for a large macro site aggregating multiple 3G & LTE cells/carrier can easily go in several Gbps which most of currently deployed backhaul solutions cannot support.
  • Performance measure including latency, jitter and frame loss thresholds is significantly stringent.
  • Coordination aspect of LTE-A requires phase synchronization in addition to currently supported frequency synchronization.


Solutions like Carrier Ethernet, IP/MPLS, (N)GPON are developing features to support these requirements but the economics of increased costs because of these features is still not very clear.

Multiple Categories of Small Cells require different backhaul strategies

Small Cells working as a component of Heterogeneous Network (HetNet) are being lauded as a solution to increase capacity within the constraints of available spectrum. Small Cells have been around for quite some time and have been deployed to achieve different purposes.


  1. Increasing coverage using Femto cells for 2G/3G/4G and IBS (In-building solutions) using DAS (Distributed Antenna Systems) for areas of bad coverage.

Utilizing fixed broadband access is a common and viable solution here.

  1. Increasing Capacity and Offloading Macro sites:
    • Carrier Wi-Fi offload using 802.11n/ac
    • Use of Metro / Micro / Pico Cells

Depending upon available and planned capacity, either a dedicated backhaul solution can be built or existing fixed broadband can be used.

  1. Increasing Capacity and enhancing Customer Experience using:
    • Seamless Carrier Wi-Fi with 802.11u & Hotspot 2.0
    • Coordinated small cells with Macro via eICIC or CoMP

This category of small cells requires special attention for backhaul planning because of requirement of similar features as mentioned for LTE-A above.

How Cloud RAN impacts the backhaul networks

Cloud RAN (radio access network) is tipped as the disruptive technology which will change the status quo of mobile industry. Enabled by 3GPP distributed architecture of BBU (baseband unit) and RRU (remote radio unit), Cloud RAN starts with BBU centralization to aggregate multiple RRUs with the benefit of resource sharing and increased utilization.


This distribution introduces a new category of fronthaul which is being standardized on CPRI (common packet radio interface). CPRI is a standard developed with cooperation of all major mobile network vendors including Ericsson, Huawei, NSN, Alcatel-Lucent and NEC. It transports raw and uncompressed data between BBU and RRU and requires very high performance characteristics, e.g. latency of 5μs, jitter of ±0.002 ppm and Round Trip Delay Accuracy of ± 16.276ns.


Examples of bit-rates required for a single cell. One site can have several cells/RRUs depending upon configuration.

Most of the current backhaul technologies cannot meet these performance criteria or the exponential bandwidth per site.

Alternatives for CPRI front-haul

  • Dark Fiber is a viable solution only if plenty of fiber resources are available for each cell site.
  • DWDM/OTN can be a suitable solution for macro sites where multiple RRU (CPRIs) can be aggregated to offer high scalability. But it introduced very high cost and negates the CAPEX & OPEX benefits gained by BBU centralization.
  • CWDM is significantly less expensive than DWDM/OTN however it has limited scalability, OAN and protection options.
  • Ethernet (Compressed) with 1:2 or 1:4 ratio makes it viable to transport CPRI over Carrier Ethernet like solutions, however it’s not proven yet whether it will meet the performance specs of CPRI
  • Radio (Compressed/uncompressed) over millimeter radio is being developed as a viable alternative of wired solutions which is attractive for shorter distances and distributed small cells who are coordinated with each other via BBU centralization.
  • WDMPON/NGPON2 are potential solutions to be deployed in point-to-multipoint architecture which utilized the existing (G/E)PON deployments. Important concern is cost economics for WDMPON and solution maturity for NGPON2.

To summarize, the future mobile backhaul network should provide reduced complexity and OPEX. They have to support for legacy (2G) as well as the bandwidth scalability for LTE-Advanced and small cells. Efficient and cost effective support of CPRI fronthaul will be the major inhibitor in adoption of Cloud-RAN.

Network Cloud brings a ray of hope for stagnating Telecom industry

network plus cloud

Telecom was once seen as innovative and high growth industry, telecom companies were at the forefront of the first high-tech bubble during late 90’s and early 2000’s. Traditional telecom services of voice, messaging, video broadcast and data connectivity were profitable and operator were only competing with local rivals who were all protected by their respective regulatory frameworks. As a result, they became a bit too comfortable in their modus operandi and didn’t feel the urge to innovate further.

But, other Internet OTT (over-the-top) companies exploited the lack of innovation by telecom and kept introducing services with not only enhanced features but also global/ubiquitous reach and new business models. The result was emergence of other alternatives for multimedia applications to run over data networks and even replace the traditional telecom services. As a result, telecom operators find themselves as a mere connectivity provider which is very similar to other utilities like water and electricity.

This new role could be fine if it provided a sustainable business model but with cloud, mobility and social media, telcos are facing an explosion of data on their networks where they have to keep on investing to expand the capacity but resulting revenues are not growing in-line with the expenses. Result is reduced profitability and excessive pressure on both CAPEX and OPEX spending.

network cloud

Now the rise of Network Cloud enabled by SDN (software defined networking) and NFV (network functions virtualization) has promised to address several of the challenges that exist for Telecom Operators worldwide and a major factor as why they are less competitive than OTT players.

Both SDN and NFV have taken the telecom industry by storm in the past two years. The level of excitement, buzz and industry activity during this time is unseen in telecom history. Perhaps this is as major a paradigm shift as moving from analogue to digital or from TDM to IP.

The industry has seen a rapid development in SDN/NFV technology solutions. The incumbent vendors are busy in evolving their SDN and NFV stories while it has become a lot easier than before for new players to not only enter the market but to pose a head-on challenge to incumbents in major network build-outs as was shown by AT&T during its selection of vendors for Domain 2.0: User-Defined Network Cloud.

It’s important to understand the benefits brought by network cloud in order to analyze whether the unprecedented hype surrounding SDN and NFV is justified or not.

Reduced disparity between costs and revenues

Using COTS servers and white label switches will have a direct impact on both CAPEX and OPEX due to shared hardware resources, reduced space, power and cooling requirements etc. When we couple the reduced costs with the ability to offer more innovative services (e.g. bandwidth on demand for enterprise applications or flexible, reconfigurable, virtualized data-centric network), the disparity between costs and revenues will decrease.

Increased flexibility, agility and resource utilization

Currently network resources are statically configured and cannot be moved when & where needed. The networks are planned and dimensioned for peak capacity which results in wasted infrastructure resources.

As with Cloud Computing, virtualizing network resources will enable dynamic allocation and hence dimensioning for average rather than peak.

Faster time to market (TTM)

Current telecom product development cycles are hardware-oriented which normally run from idea generation, development of business plan, definition of requirements, requests for proposals, evaluations/negotiations, shipment, deployment, typically complex OSS/BSS integration, acceptance testing and finally service introduction. The result is that, on average, it takes 12 – 24 months to introduce new service for telecom operators. While they are competing with OTT players who, by relying solely on software solutions and Cloud technologies, are able to bring a new service in less than 3 months. Network virtualization shall enable the same kind of TTM to that operators can respond must faster to changing market needs.

Reduced Complexity and easier inventory management

Today’s networks have enormous variety of proprietary hardware appliances which mandated operators to follow hardware obsolescence cycles to ensure ROI (return on investment). Keeping & tracking inventory & spares for all different kinds of hardware is also a major overhead in network maintenance.

By utilizing industry standard COTS (commercial off-the-shelf) servers and white label switches, the operators will be able to significantly reduce the type of hardware to be managed.