Myths and realities of Distributed Access Architectures for cable
Distributed Access Architectures (DAA) for cable are a hot topic. They hold many promises of efficiencies and performance improvements to keep cable networks in good stead. They are far more agile, flexible, and lighter users of power.
When building a DAA related strategy, however, care has to be taken to understand the realities of a DAA deployment. DAA has architectural options and is also associated with the virtualization of CCAP. As manufacturers move toward commercializing DAA solutions, confusion abounds as to the pros and cons of these architectural options. In a world of roadmap only solutions from some vendors full of technological myths, here are a few realities to aid your strategy.
Let’s start with a definition. A distributed access architecture is where the functional components of a legacy centralized CMTS/CCAP architecture such as Controller, DOCSIS MAC and DOCSIS PHY are migrated to different locations of the end to end network to give benefits as networks evolve. At present, there are two predominant approaches to a DAA:
- Remote PHY (R-PHY) where the DOCSIS MAC stays at the hub-site and only the PHY is moved to the fiber node
- Remote MAC/PHY (R-MAC/PHY) where both the MAC and the PHY are pushed down to the node.
DAAs are not new. In 2013, CableLabs published annexes to the DOCSIS 3.0 family of specifications to account for modifications for deployments in China, although not limited to use in China. These specifications, known as C-DOCSIS, are for data only (no DVB video) and embrace the concepts of R-PHY and R-MAC/PHY. First data only devices were made by Sumavision, ZTE, but now we see most incumbent vendors now offering R-PHY type node devices.
Current R-PHY is an evolution of the CableLabs modular CMTS standards where DOCSIS MAC and PHY functions were distributed within a centralized headend site and delivered using standard HFC distribution to fiber nodes.
At Nokia, we talk about Virtual CCAP (vCCAP). This is an enhancement of the above options by also distributing the CCAP/CMTS controller back to a central datacenter, thus eliminating a physical CCAP/CMTS device in the hub.
Figure: schematic representations of centralized CCAP, DAA options for R-PHY and vCCAP (R-MAC/PHY)
So is C-DOCSIS the same as vCCAP?
Actually, there are numerous and big differences between C-DOCSIS and Virtual CCAP.
- C-DOCSIS is data only. It does not support legacy video or set-top box control. As a result, it can only be deployed as an overlay on the existing HFC network
- Today’s C-DOCSIS solutions are limited to 16 channels, moving to 32 channels and eventually D3.1. But Nokia’s vCCAP supports Full Spectrum DOCSIS today, lighting up the entire spectrum and delivering greater than 6 Gbps per coax
- C-DOCSIS solutions require operators to manage and provision each node as an individual Layer 2 CMTS. This dramatically increases the operational burden for operators by changing the paradigm from Layer 3 to Layer 2 and exploding the number of managed devices in the network. In the vCCAP architecture, each node is seen as just a line card in a massively scalable, distributed Layer 3 CCAP. There is no change in how the network perceives the CCAP system and how the MSOs would operationalize it.
As a result, C-DOCSIS doesn’t provide the investment protection or growth potential of a vCCAP strategy.
An overlay for video traffic
Before we go on to discuss DAA architectural options, I wish to highlight a couple of points about Modulation Error Ratio (MER) and video overlays in downstream. Field testing of DAA actually shows a MER improvement due to PHY being in the node — an important consideration when planning to deliver DOCSIS3.1 in both downstream and upstream. This is great news for data only systems, but let’s not forget that until cable operators go ALL-IP, they will wish to deliver a combined DVB & DOCSIS service. This requires an RF overlay architecture in the node that combines RF signals. Operators should be aware that whenever two RF signals are combined there is a measurable degradation of the MER of the combined services. It can be concluded from this that strategies leveraging DAA architectures should optimally deliver all services without overlay technologies for best performance.
The Remote-PHY Vs Remote-MAC/PHY debate
Central to a lot of the current confusion and fueling many myths as vendors position their particular flavor of DAA are the differences between R-PHY and R-MAC/PHY. Is one option better than the other in terms of performance, power consumption and conformance to standards?
Well, they are both valid architectural options, conforming to CableLabs specifications. One of the main differences however is that current R-PHY architectures require the DOCSIS MAC function to reside in the hub location (CMTS or CCAP). This results in a higher level of equipment — i.e., “Big Iron” in the hub.
R-PHY is an interesting solution for incumbent CMTS/CCAP vendors motivated to protect their installed base and revenue stream. Moving the PHY to the node does provide some cost and performance benefits by replacing analog optics with digital optics. However, it does not fully capitalize on advances in modern silicon and software to unleash maximum benefits for MSOs, and it has several shortcomings:
- It unnecessarily separates the DOCSIS MAC and PHY, which creates timing synchronization issues among others
- It leaves the CMTS/CCAP core, which is a redundant router, sitting right next to the existing edge routers
- It forces MSOs to keep a large physical box (i.e., “Big Iron”) in the headend/hub and as such, does not fully address the cost, power and space issues faced by operators today
- It limits operators’ ability to consolidate hubs given the tight timing requirements from hub to node.
Furthermore, as the industry moves towards DOCSIS full duplex working, we understand that R-MAC/PHY will be preferred at the fiber node site for correct functionality. Full Duplex DOCSIS will likely require advanced scheduling in frequency, timing and power domains which may only be delivered from a collocated MAC/PHY function.
A reality of deploying DAAs is that the fiber node site will have to house more complex equipment than today’s existing fiber node. This in turn implies a higher power consumption — so does R-PHY fit into an existing node site power budget and R-MAC/PHY not?
Not at all, especially when requiring video and data capabilities. Understandably it is hard for operators to compare vendors’ nodes as many are not yet in commercial production. Furthermore, the power consumption of each one will vary on its particular specification — big, small, data only, video and data, RF overlay, one service group, two service groups, RF launch power etc. Cable operators have to compare apples with apples.
In reality, both R-PHY and R-MAC/PHY nodes consist of an Ethernet receiver, a processing complex, PHY and RF amplification stages. If you think about it, most node power consumption goes into the required RF amplifiers for both options.
The situations where an R-PHY node would consume significantly less power than a R-MAC/PHY node is typically where there are substantial differences in the configuration of the R-PHY and R-MAC/PHY node with respect to number of serving groups, RF launch power and overlay capabilities.
So if power is not a concern, what about the differences in complexity and risk?
The incremental complexity between an R-PHY node and an R-MAC/PHY node is very small. All the node components will be designed to have greater than 100k hours mean time before failure. Both options should have similar reliability. Some may argue however that R-PHY is more complex. This is because splitting the DOCSIS MAC and PHY creates more complexity with the additional protocols that are required along with the need for time synchronization and scheduling.
In this sense R-MAC/PHY is certainly not a new concept in networking and telecommunications. Think about other common access technologies such as Wi-Fi, LTE, DSL, etc. They keep the MAC and PHY together while distributing them as close to the end-user as possible. It is a proven concept.
So finally, let’s compare the two options in terms of standards. Both are standardized architectures. When leveraging R-MAC/PHY, the standardized MAC/PHY function is just moved to the node location. Furthermore, backhaul between node and hub is performed with standardized 10G Ethernet & routing components.
To dig in a little further with the comparison, actually in the data plane, R-PHY is more proprietary than R-MAC/PHY.
- R-PHY: requires the use of a variety of specifications (DEPI / UEPI/ GCP/ R-DTI / R-OOB), which are deployed only in the cable sector. Router and switch vendors do not implementing these. They are not vendor proprietary, but are cable industry proprietary. It also has to be set up as a direct connection from hub to node (just like the current architecture). It can’t be carried over “any architecture” without careful consideration of tunnel scaling and timing distribution requirements.
- R-MAC/PHY leverages standard 10G Ethernet. This is the same as any switching/routing device in the world. In fact, standardized Metro Ethernet Forum services can actually be used for this backhaul.
A lot of the information contained in this article is gleaned from customer conversations as well as field testing of our Unified Cable Access solution. You may know that Nokia as a leading global technology supplier has world leading experience in the concepts required for successfully deploying cable Distributed Access Architectures and virtual CCAP.
As MSOs evolve their access strategies and manufacturers move toward commercializing DAA solutions, confusion abounds as to the pros and cons of the architectural options. In a world of roadmap only solutions from some vendors, we hope to have answered some of the questions you may have about DAA and dispelled some of the technological myths surrounding R-MAC and R-MAC/PHY.