HE virtualization – a path to the headend of tomorrow
Date of publication: 14.12.2017
An existing CCAP contains a physical-layer (PHY), downstream QAM/OFDM modulators and upstream QAM/OFDM demodulators, DOCSIS MAC functions as well as some routing functions. Physical QAM/OFDM signals require dedicated hardware such as ASICs/FPGAs, making a CCAP sophisticated and expensive box; so-called “Big Iron”.
To meet growing bandwidth and traffic demands, cable companies prepare to roll out DOCSIS 3.1, and progress in the further segmentation of their networks. This requires operators to install more “Big Iron” in the headend. But with space running out in centralized headends (hubs) that are already fully packed with power-hungry, dedicated systems, operators will face a new challenge.
With a Remote PHY, it is possible to split the physical signals from the CCAP and push QAM and OFDM modulation out to the Remote PHY Devices (RPDs) while CCAP Core routing can run out of the larger hubs.
Thanks to the separation of the core and the PHY, virtualization of the core functions is possible. MSOs gets the option to migrate from highly specialized hardware for their CCAP functions to run those functions on commercial, off-the-shelf IT platforms (x86 servers). Such architecture is easily expandable and can be upgraded to new delivery standards as needed.
Once the Remote PHY Devices (RPDs) are out of the hub, the virtualized CCAP Core can run on the same hardware infrastructure that is used for the vitalization of all the headend functions:
- OOB – Cable operators install many supporting systems to manage their HFC networks. Many of these solutions are based on dedicated hardware and require a manual connection between the system and the service group that requires a network management, monitoring and troubleshooting. With virtualization, the OOB system, which we call the OOB Core, can operate on COTS servers, allowing operators to eliminate dedicated hardware in headend/hub, and as a result, save some space and power. A new, virtualized OOB system will be pure software that receives a sampled signal from hundreds of RPDs (delivered by the NDF/NDR function of Remote PHY－ see the OOB article). In this case, there is a need to increase the scalability of the OOB systems. It is a simple process of instantiating another copy of the software that operates on the same resource as the rest of the functions.
- VIDEO – By pushing the signal modulation into the RPD, all video functions, such as de-jittering, are simplified and can be performed by pure software, which we call the VIDEO Core. A complicated function like PCR restamping, multiplexing or scrambling can be done purely by software. Additionally, as video transport is handled by IP Multicast, it could be easily replicated on switches and routers, which are already used for the rest of the services such as DATA and OOB (see the unified network article). This allows for further simplification of the headend infrastructure, and even the centralization of all video functions into one, central headend.
- DATA – The holy grail of virtualization is the full virtualization of DOCSIS MAC, which allows a scalable environment to build fast and reliable, high-speed Internet services. With virtualization of the DATA core, it is possible for MSO to gain the flexibility of a centralized or decentralized architecture. Due to virtualization and distribution, an RPD operator can easily and efficiently design some high-availability into all of the RPDs in their network with extensive resources as needed as was done in case of CMTS/I-CCAP – RF switches, redundant line cards and a complicated set of coaxial cables. A virtual data core allows for the maximal consolidation of the headends with the possible migration of all data/control planes and DOCSIS MACs to central Data Centre, and a reduction in the number of hubs. Also, the DATA Core functions will operate on would run on COTS servers, for which performance is increasing each year, as opposed to Big Iron platforms which are seeing major performance upgrades in 3-5 year intervals. Such a scenario allows for the steady growth of operators’ infrastructure without the need for big, expensive upgrades of the CCAP Core infrastructure.
Regardless of whether an operator is going to virtualize single functions or all the headend functions, headend virtualization will allow MSOs to reduce OPEX and CAPEX, thanks to:
- The usage of the COTS servers that are standardized platforms which price is constantly dropping.
- Full utilization of server platforms thanks to sharing computing resources between different virtual functions such as OOB, VIDEO and DATA.
A reduction in the size of the hubs or a reduction in the number of hubs with centralization of the CCAP Core function into one or two Data Centres.
Want to read more about challenges and opportunities for MSO with Remote PHY implementation?