 |
Debunking NBN Myths
Wednesday, April 22, 2009, 01:55 AM
IP Broadband Network Architecture Considerations by George R J Green
In building any communication network it is usually about trade offs and compromises between performance, services offered and costs and what are the market/user requirements to be supported. Then working out what is the most efficient transport mechanism to use for such applications. And it depends if you are starting from an existing system, Cable TV or Telco, or if you have a clean slate to start from scratch, a green field system.
Both Cable TV and Telco systems evolved having to consider legacy issues along the way which has to some extent limited their full capabilities in one form or another. Cable TV has been a very efficient way of distributing multi-channel video services and supports data and VoIP services by using DOCSIS over the existing cable plant. However, Cable is still basically a broadcast type network carrying MPEG based video channels in a broadcast mode along with IP data and VoIP services in a communications mode and is not a true IP broadband communications network. Even the latest DOCSIS 3.0 is limited to 160 Mbps downstream and 120 Mbps upstream when they bond four channels together. However, as Professor Rod Tucker mentioned in his previous paper, DOCSIS is still a shared medium between multiple users so the bits per user are considerably less. This means that DOCSIS 3.0 is more likely to be competitive with xDSL than FTTP systems, even when DOCSIS goes to 400 Mbps.
The Telco networks were designed to carry circuit switched telephony and so their natural evolution path has been to migrate to xDSL to provide data services and perhaps somewhat reluctantly VoIP services. However, Telcos have now successfully proven, as in the case of PCCW in Hong Kong who now has more video subscribers than the I Cable, that they are very capable of providing IPTV video services. The main challenge that Telcos have is in supporting High Definition (HD) video services and in particular delivering multiple HD video channels to each user so as to support different programs to multiple televisions in a home. In the case of PCCW and some others they already offer a 30 Mbps service and so they can support multiple HD channels now along with internet data and VoIP services simultaneously. In the meantime Telco legacy issues mean that they must still carry and support circuit switched traffic over their copper lines.
The more important question is if you are not a Cable TV company or a Telco then what would you build today. That is, not having to worry about legacy issues is the ideal position to be in during the network architecture decision making phase and so you can focus on what is the most efficient means to transport the services based on what the market demands and user requirements are. To a large extent the market demands are still basically the same, multiple HD videos per user, VOD or real time streaming, internet data and VoIP. Services such as video conferencing are easy to support since they are low bandwidth traffic and in the noise almost in comparison to the downstream HD video traffic requirements which is the real bandwidth hog in the network. The other critical network aspect is in being able to support high value commercial users on the same communication platform.
It’s intuitively obvious that if you are to build a new network from scratch today you would not build either a Cable TV network or a copper based Telco network. Any new network being built from scratch today would be an Ethernet communications network since it is the most efficient means of transporting IP packets and today the world is all Ethernet based.
At this time the most logical and lowest cost network architecture choice would be to build an IEEE 802.3ah GEPON/EPON FTTP network. (Gigabit Ethernet Passive Optical Network, also known as Ethernet PON, EPON). GEPON supports a symmetrical (downstream and upstream) 1.25 Gbps. GEPON's biggest advantage is its use of native Ethernet protocols combined with the fact that GEPON systems are the lowest cost and most widely deployed fibre technology to date. China alone, as of the end of 2008, has deployed 20 million subscribers using GEPON and Japan has deployed the same technology. In fact, Asia leads the world in GEPON deployments. This is because the fibre cable and the GEPON terminal equipment are very low cost and the hardware costs get lower by the day. However, the real cost, around 80%, off an FTTP network is in the actual construction of the fiber network and not the technology used.
GEPON provides seamless connectivity for any type of IP-based or other packetized (Ethernet) communications as it employs a single Layer 2 network that uses IP to carry data, voice, and video. Since Ethernet devices are present everywhere from the home network to regional, national and global backbone networks, implementation of EPONs has proved to be the most cost-effective. It also provides secure communication as encryption is provided at both ends i.e. upstream and downstream, so the chances of eavesdropping are minimal. As the transfer rates in Ethernet scale up by the day the scalability of GEPON is limitless.
An alternative, mentioned by Professor Rod Tucker and introduced in 2003 is the ITU-T G.984, GPON which uses a new native Generic Encapsulation Method (GEM) transport layer that supports multiple "non-native" transport protocols including ATM, Ethernet, and TDM. The original intention here was to broaden support and market acceptance by supporting multiple protocols, but the effect has been to add complexity to those systems not requiring this additional protocol support.
A key characteristic of GPON is the 2.5 Gbps downstream data rate and the 1.25 Gbps upstream data rate. However, GPON operates in a very similar fashion to GEPON when supporting Ethernet as its primary transport protocol since Ethernet, gigabit Ethernet and 10 Gb Ethernet do not support a 2.5 GHz clock rate, unlike GEPON and hence GPON does not benefit from the availability of low-cost Ethernet optical components currently used in GEPON systems.
Lacking the economies of scale and with the specifications still being updated, GPON deployment has lagged behind that of GEPON. Nonetheless, with the backing of FSAN, the Full Service Access Networks organization, a group of worldwide carriers who play an important role in the ITU standards body, and the prospect of simplifying the protocol support to focus on Ethernet only, GPON has garnered a great deal of interest in North America and other geographic regions seeking to adopt a gigabit PON standard for FTTP deployments and so GPON may still have its day in the sun yet.
Regardless of which solution is used for FTTP, GPON or GEPON the question becomes what data rate do you need/want at the user end and so depending on whether you take GPON at 2.5 Gbps or GEPON at 1.25 Gbps it depends on how many times you split the fibre from the OLT (Optical Line Termination) unit to the home ONU’s (Optical Network Unit). This can be 4, 8, 16 or 32 so pick a number.
If you use GEPON at 1.25 Gbps and split the fibre by 16 then you end up with a dedicated bits/per/user of 78.125 Mbps which is in fact more than enough for most consumer applications and beats all of the existing competitors. In reality the fibre ONU operates up to 1.25 Gbps on the fiber side and up to 100 Mbps on the user LAN side so in effect you actually have a 100 Mbps system. Data is distributed on a dynamic and statistical basis and not everyone is hogging all the bandwidth at any one given time and so saying its only 78 Mbps is just semantics, as a user you would have network access at the full 100 Mbps rate.
Note that 1.25 Gbps is the line rate and the actual IP data throughput for Ethernet with GEPON is closer to 1 Gbps and if you account for overhead it will be even less, around 900-950 Mbps downstream and 700-800 Mbps upstream. However, 2.5GEPON is now available offering a 2.5 Gbps line rate downstream and 1.25 upstream. Regardless and as mentioned above the actual IP data throughput you get/need depends on the fibre split used.
In the case of an HD video stream it will run between 8 Mbps to 12 Mbps so such an FTTP network can easily support multiple HD streams running into a home simultaneously. On average you could assume 10 Mbps per video stream and since most ONU’s have four Ethernet ports then 40 Mbps with four simultaneous videos streams there is more than enough capacity for any one home. The other advantage of running an IP network is that you can make all or the majority of the IP video streams available as on-demand titles using IP multi-tasking. This enables you to deliver hundreds of on-demand titles simultaneously across your IP network with minimal bandwidth.
IP Multitasking using IGMPv3 (Internet Group Management Protocol) is the most efficient way to network IP video services. It reduces the network bandwidth requirements by more 95% and reduces the number of video servers by the same amount versus the current unicast VOD systems used in Cable TV today. In addition, there is no network dimensioning required when using multicast and so you never need to worry about scalability and running out of bandwidth on popular videos as can happen with today’s unicast VOD systems. After all, is this not what consumers want, content on demand? Plus with an IP network you eliminate all the need for headend equipment as is used in Cable TV networks such as QAM modulators. If required you can simultaneously run some unicast streams but 95% of the user requirements is most likely going to be in favour of on-demand content. This also makes the set top terminal equipment simple, low cost and standardized if all content is on-demand and IP based.
If you decide that you need more than 78 Mbps per user then you can either reduce the fibre split count or in the future change out the fiber OLT and ONU terminal equipment, as was pointed out by Professor Rod Tucker in a previous article, to be whatever since the costs will not be that significant. Plus users can be charged based on the connect rate they subscribe to.
In the case of commercial users they can have any data rate up to the full 1.25 Gbps into their LAN network if required. And since the network is all Ethernet based it provides seamless access for all commercial applications, local, national or global, as well as support any number of VPN applications.
The reality is that the existing cable systems in Australia have had their day. Cable systems globally are all moving as fast as they can towards becoming 100% IP networks. As C-COR Broadband's Dermot Cox already stated, “in new builds cable systems are building FTTP”. The fact that cable systems have legacy issues or want to use technologies like the proposed RFoG technology (Radio Frequency over Glass) to get there does not mean you have to use such network architectures. If cable systems want to continue delivering hundreds of broadcast type RF channels to each user in this manner then so be it.
Most new systems being built from scratch today are building either 100% fiber FTTP networks or FTTC with xDSL and not Cable TV type architectures and as noted above even Cable TV is moving towards FTTP. In the case of Telcos they have to decide how far they want to upgrade, to FTTP or FTTC with xDSL to remain competitive. In the case of Telstra, if they want to compete with NBN they will either have to move towards an FTTP network on the Telco side of their business or significantly upgrade their Cable system but why have two parallel systems offering the same services, it’s very expensive and becomes redundant at that stage.
As an example a FTTP 10GEPON IP multicast network architecture is illustrated below. This is much more bandwidth efficient than a unicast system. It uses 95% less bandwidth and video servers and can simultaneously deliver hundreds of HD on-demand video titles and real time HD streams. In addition, this network architecture can seamlessly support any commercial application.
The future looks very bright for FTTP and fibre with ever increasing bandwidth capabilities. 2.5GEPON is already available providing 2.5 Gbps per fibre and 10GEPON a new 10 Gbps standard should be approved in Sept. 2009, IEEE 802.3av, offering either symmetrical 10 Gbps per fibre or asymmetrical offering 1 Gbps upstream. In addition, you have 20 Gbps WDM PON Ethernet being launched now providing 1.25 Gbps per user and a planned 80 Gbps WDM providing 2.5 Gbps per user that will be available within the next two years. Regardless and as mentioned previously there is always a trade off against costs and selecting the most efficient architecture based on the services to be provided. Commercial applications are going to be as equally demanding of bandwidth with LTE coming on-stream demanding 300 Mbps per cell tower. The reality of FTTP is if you build it they will come.
George R J Green
President and CEO
Pacific Broadband Inc.
Level 3 Three Pacific Place
1 Queen’s Road East,
Hong Kong
Tel: +852-6278-1906
Fax: +852-2855-6811
Email: g.green@pacific-broadband.com
 [comment]
|
|
| |
| Comments |
| No
Comments |
|
|
|
[print]
[email]
