Getting your Trinity Audio player ready...
|
Fragmented and Costly Networking Topologies and Operations
Of the 88 network operators I’ve heard from over the last year, only 8 report having a “single vendor” network; the average is three. None say their networks are made up of a single technology; here four different technologies are average. The devices deployed range in age from less than six months to over six years, they involve an average of almost three distinct management systems and operations specializations. This diversity is created, say operators, by the inherent distributed and hierarchical nature of networks, the mixture of services created, and the fact that network vendors have for years employed specialized technology for each layer of network infrastructure.
So what? Well, operators on the average say that this distributed network diversity costs them 20% more than necessary:
- 3x the overall service outage minutes
- 33% increased power and cooling costs
- 4x the time they have to devote to staff training
All of this, at a time when operators worldwide are struggling to deliver improved profits in an era where the normal sources of profit improvement, increased average revenue per user (ARPU) and total addressable market (TAM) are elusive, to say the least. If you can’t increase revenue enough to meet investor expectations, then costs need to be lowered.
This situation is completely understandable, having been the rule in networking for decades, but it’s become intolerable. To resolve the problem, we need to take a more critical look at the network – do we need it all?
- Yes, we need it to be distributed, but can we use fewer components for that?
- Do we need all the locations or can we merge them?
- How can we standardize some of the components – streamline the hardware to streamline operations?
Or in short – can we optimize the network footprint? To do so, we need to add another “D” to the network’s properties of distributed and diverse, the property of disaggregation.
Network Transformation Goals
How can the third “D”, help us here? The answer is simple; by reducing the impacts of the other two. We can’t eliminate distribution in networks; they’re distributed because the customers are, but we can reduce the impact of this distribution by leveraging cloud principles to improve networks, creating pools of standard resources, and basing functionality on software. Disaggregation then also offers us a real opportunity to eliminate those diverse, wasteful, purpose-built devices that require their own set of spares and introduce different operations practices and operations complexity. It may seem like a contradiction, but the disaggregation of network elements facilitates the convergence of both network hardware and operations.
You surely have seen many network diagrams over the year, and if you think back on each, you’ll realize that the equipment is always represented by a box…of some shape or another, perhaps, but a box in all cases. What that says is that infrastructure is first and foremost a set of functional roles connected by transport facilities. Why did we specialize this, so a square, rectangular, trapezoidal, or whatever shape of box we drew is necessarily implemented with a totally different sort of device, and why did we set it up so that an increase in capacity or availability always meant more boxes? Because we build networks today the way we built data centers half-a-century ago. You don’t have a different cloud, or even a different cloud service, for every application, right? Why not make the cloud model the standard model for networks, then? And like the cloud is being transformed by automation in general, and AI in particular, the hardware and operations convergence that our new disaggregated model enables transforms automated NetOps and facilitates AIOps.
To do all these good things, we have to start thinking of those boxes in a different, functional, disaggregated, way = as software functions. The box becomes less relevant. For years, the telecom industry has focused on optimizing the box to lower the cost, and we ended up increasing the OpEx of maintaining it. The disaggregated hardware model means every network feature can be implemented on a common platform, one that creates a pool of resources just as the cloud does. The disaggregation of hardware and software enables a converged operations model like the one created by DevOps or orchestration tools in the cloud.
Benefits of Network Convergence
Suppose that network operations operate in the same way on all the elements, whether they’re deployed singly at the edge or deeper as a pool. Now you have a common operations model across all infrastructure elements. If each hardware element you use is the same “white box” type, then you need only one set of spares. You have one operations skill set, one team, one training process, complete and uniform management visibility and practices. Operations automation is easy with this approach, because the vendor- and device-specific silos found in most of today’s networks are gone. The “single source of truth” that’s we hear so much about these days is critical to operations automation, and it’s inherent in the converged operations model that disaggregation facilitates.
OpEx, cost of spares, facility, and power/cooling costs are all proportional to complexity driven by device diversity, and disaggregation and cloud principles reduce complexity by eliminating devices, locations, and entire layers, and by automating responses to abnormal conditions to reduce human effort and errors. The ability of a pool of our disaggregated elements to increase its resources at a pool level through the addition of elements means that resilience, scalability, and availability are all optimized just as they are in the cloud.
Comcast Network Transformation with DriveNets
DriveNets Network Cloud provides an open, software-based, and converged solution for service providers. It enables hyperscaler-like networks with a unified hardware and software platform, simplifying the network architecture and reducing its footprint. With the Network Cloud, features and even device missions and layers are defined by software, so any compatible hardware device can be virtually any box, play any role.
Comcast, a major cable provider, adopted DriveNets for its Janus project, a vast infrastructure modernization initiative that exploits the principles of disaggregation and virtualization, hardware and operational convergence, and operations automation that exploits AI. Comcast’s network is the largest in the US, serving over 63 million locations. Like most broadband operators, they’ve experienced explosive traffic growth, largely due to the popularity of the live streaming of sports events, causing data traffic patterns to fluctuate massively on a daily basis. In this climate, it’s critical to capital and operational costs, improve service quality of experience, and quickly remediate faults. They found the traditional “bookended” hardware components and device-specific software and operations framework difficult to improve, and so Janus was launched to create a truly cloud-centric network model, shifting Comcast’s management and control of its core routing, switching and transport network functions to its edge cloud platforms.
Janus builds on DriveNets’ common-element white-box model that disaggregates and distributes the hosting of cloud-native network feature components, converging them into what’s effectively a set of resource pools on which once-monolithic device functionality is then hosted. This lets Comcast use AI/ML to analyze network conditions and trigger automated responses. Janus directs functionality and capacity to wherever it’s needed, in the amount needed, and at the cost essential to providing both profitable and affordable service.
Janus is the modern cloud-like, network infrastructure, and to quote Comcast’s Janus press release, “Comcast is using Network Cloud software developed by DriveNets, a leader in cloud network technology….” Their application is a perfect example of how cloud principles and disaggregation can distribute hardware and software-based network functionality to control the diversity and disorder created by the legacy device model.
Whatever services are created “over the top” of network infrastructure, the infrastructure and its connection services have to be profitable. That starts with making basic connectivity less costly, and extends to tuning service features to optimally fit customer experience requirements. A network built on a common set of elements, assembled to serve every mission but retaining their common, converged and automated operations model, is inherently better at both assessing service quality needs and creating new features to meet them. That’s exactly what DriveNets Network Cloud delivers.
Download
Which Network Architecture Is Right for You