Guest Post: UFB for Dummies
Steve Biddle likes to describe himself as a former trolley boy but nobody believes him about that (it’s true, I swear) so we’ll just call him a network engineer with a passion for explaining things simply.
Steve posted this to his blog over on Geekzone but kindly allowed me to republish it here.
Unless you’ve been living on another planet you’ll be aware that New Zealand is currently in the process of deploying a nationwide Fibre To The Home (FTTH) network. This network is being supported by the New Zealand Government to the tune of roughly NZ$1.5 billion over the next 10 years and is being managed by Crown Fibre Holdings (CFH). Work is presently underway deploying fibre nationwide, with several thousand homes now connected to this new network.
Much has been made of UFB retail pricing, and for many individuals and businesses the price they will pay for a UFB fibre connection could be significantly cheaper than existing copper or fibre connections. What does need to be understood however is the differences between fibre connection types, and pricing structures for these different services. There have been a number of public discussions in recent months (including at Nethui in July) where a number of comments made by people show a level of ignorance, both at a business and technical level, of exactly how fibre services are delivered, dimensioned, and the actual costs of providing a service.
So why is UFB pricing significantly cheaper than some current fibre pricing? The answer is pretty simple – it’s all about the network architecture, bandwidth requirements and the Committed Information Rate (CIR). CIR is a figure representing the actual guaranteed bandwidth per customer, something we’ll a talk lot about later. First however, we need a quick lesson on network architecture.
Current large scale fibre networks from the likes of Chorus, FX Networks, Citylink and Vector (just to name a few) are typically all Point-to-Point networks. This means the physical fibre connection to the Optical Network Terminal (ONT) on your premises is a dedicated fibre optic cable connected directly back to a single fibre port in an aggregation switch. Point-to-point architecture is similar to existing copper phone networks throughout the world, where the copper pair running to your house is dedicated connection between your premises and the local cabinet or exchange, and is used only by you. Because the fibre is only used by a single customer the speed can be guaranteed and will typically be dimensioned for a fixed speed, ie if you pay for a 100Mbps connection your connection will be provisioned with a 100Mbps CIR and this speed will be achieved 24/7 over the physical fibre connection (but once it leaves the fibre access network it is of course up to your ISP to guarantee speeds). Speeds of up to 10 Gb/s can easily be delivered over a Point-to-Point fibre connection.
The core architecture of the UFB project is Gigabit Passive Optical Network (GPON). Rather than a fibre port in the Optical Line Terminal (OLT) being dedicated to a single customer, the single fibre from the port is split using a passive optical splitter so it’s capable of serving multiple customers . GPON architecture typically involves the use of 12, 24 or 32 way splitters between the OLT and the customers ONT on their premises. GPON delivers aggregate bandwidth of 2.488Gb/s downstream and 1.244 Gb/s upstream shared between all the customers who are connected to it. 24 way splitters will typically be used in New Zealand, meaning that 100Mbps downstream and 50Mbps upstream can be delivered uncontended to each customer. The difference is architecture is immediately clear – rather than the expensive cost of the fibre port having to be recovered by a single customer as is the case with a Point-to-Point network, the cost is now recovered from multiple customers. The real world result of this is an immediate drop in the wholesale port cost, meaning wholesale access can now be offered at significantly cheaper price points than is possible with a Point-to-Point architecture. GPON’s shared architecture also means that costs can be lowered even further since the architecture of a shared network means dedicated bandwidth isn’t required for every customer like is is with a Point-to-Point connection. The 2.488Gbps downstream and 1.244Gbps upstream capacity of the GPON network instantly becomes a shared resource meaning lower costs, but it can also mean a lower quality connection compared to a Point-to-Point fibre connection.
Now that we’ve covered the basics of architecture we now need to learn the basics of bandwidth dimensioning. Above we learnt that a CIR is a guaranteed amount of bandwidth available over a connection. Bandwidth that isn’t guaranteed is known as an Excess Information Rate (EIR). EIR is a term to describe traffic that is best effort, with no real world guarantee of performance. The 30Mbps, 50Mbps or 100Mbps service bandwidth speeds referred to in UFB residential GPON pricing are all EIR figures, as is the norm with residential grade broadband services virtually everywhere in the world. There are is no guarantee that you will receive this EIR speed, or that the speed will not vary depending on the time of the day, or with network congestion caused by other users. With Voice Over Internet Protocol (VoIP) replacing analogue phone lines in the fibre world, guaranteed bandwidth needs to also be available to ensure that VoIP services can deliver a quality fixed line replacement. To deliver this UFB GPON residential plans also include a high priority CIR of between 2.5Mbps and 10Mbps which can be used by tagged traffic. In the real world this means that a residential GPON 100Mbps connection with a 10Mbps CIR would deliver an EIR of 100Mbps, and a guaranteed 10Mbps of bandwidth for the high priority CIR path.
Those of you paying attention would have noticed a new word in the paragraph above – tagged. If you understand very little about computer networking or the internet you probably just assume that the CIR applies to the EIR figure, and that you are guaranteed 10Mbps on your 100Mbps connection. This isn’t quite the case, as maintaining a CIR and delivering a guaranteed service for high priority applications such as voice can only be done by policing traffic classes either by 801.2p tags or VLAN’s The 802.1p standard defines 8 different classes of service ranging from 0 (lowest) to 7 (highest). For traffic to use the CIR rather than EIR bandwidth it needs to be tagged with a 802.1p value within the Ethernet header so the network knows what class the traffic belongs to. Traffic with the correct high priority 802.1p tag will travel along the high priority CIR path, and traffic that either isn’t tagged, or tagged with a value other than that specified value for the high priority path will travel along the low priority EIR path. Traffic in excess of the EIR is queued, and traffic tagged with a 802.1p high priority tag that is in excess of the CIR is discarded.
For those that aren’t technically savvy an analogy (which is similar but not entirely correct in every aspect) is to compare your connection to a motorway. Traffic volumes at different times of the day will result in varying speeds as all traffic on the motorway is best effort, in the same way EIR traffic is best effort. To deliver guaranteed throughput without delays a high priority lane exists on the motorway that delivers guaranteed speed 24/7 to those drivers who have specially marked vehicles that are permitted to use this lane.
There are probably some of you right now that are confused by the requirement for tagged traffic and two different traffic classes. The simple reality is that different Class of Service (CoS) traffic profiles are the best way to deliver a high quality end user experience and to guarantee Quality of Service (QoS) to sensitive traffic such as voice. Packet loss and jitter cause havoc for VoIP traffic, so dimensioning of a network to separate high and low priority traffic is quite simply best practice. Performance specifications exist for both traffic classes, with high priority traffic being subject to very low figures for frame delay, frame delay variation and frame loss.
UFB users on business plans also have a number of different plan options that differ quite considerably to residential plans. All plans have the ability to have Priority Code Point (PCP) transparency enabled or disabled. With PCP Transparency disabled, traffic is dimensioned based on the 802.1p tag value in the same way as residential connections are. With PCP Transparency enabled, all traffic, regardless of the 802.1p tag, will be regarded as high priority and your maximum speed will be your CIR rate. As the CIR on business plans can be upgraded right up to 100Mbps, GPON can deliver a service equivalent to the performance of a Point-to-Point fibre connection. Business users also have the option of opting for a CIR on their EIR (confused yet?). This means that a 100Mbps business connection can opt for a service bandwidth of 100Mbps featuring a 2.5Mbps high priority CIR, a 95Mbps low priority EIR, and a 2.5Mbps low priority CIR. This means that at any time 2.5Mbps will be the guaranteed CIR of the combined low priority traffic. The high priority CIR can be upgraded right up to 90Mbps, with such an offering delivering a 90Mbps high priority CIR, 7.5Mbps low priority EIR, and 2.5Mbps low priority CIR.
You’re now probably wondering about 802.p tagging of traffic. For upstream traffic this tagging can be done either by your router, or any network device or software application that supports this feature. Most VoIP hardware for example already comes preconfigured with 802.1p settings, however these will need to be configured with the required 802.1p value for the network. Downstream tagging of traffic introduces whole new set of challenges – while ISP’s can tag their own VoIP traffic for example, Skype traffic that may have travelled from the other side of the world is highly unlikely to contain a 802.1p tag that will place it in the high priority CIR path, so it will be treated as low priority EIR traffic. ISP’s aren’t going to necessarily have the ability to tag traffic as high priority unless it either originates within their network, or steps are taken to identify and tag specific external traffic, meaning that the uses of the CIR for downstream will be controlled by your ISP.
It is also worth noting that all of the speeds mentioned in this post refer only to the physical fibre connection. Once traffic leaves the handover point, known as an Ethernet Aggregation Switch (EAS) it’s up to the individual ISP to dimension backhaul and their own upstream bandwidth to support their users.
As part of their agreement with CFH, Chorus dropped their Point-to-Point fibre pricing in fibre existing areas in August 2011 to match UFB Point-to-Point pricing, which means customers currently in non UFB areas will pay exactly the same price for a Point-to-Point fibre access as they will do in a UFB area if they choose a Point-to-Point UFB connection. UFB GPON fibre plans won’t be available in existing fibre however areas until the GPON network has been deployed, either by Chorus or the LFC responsible for that area. In all UFB areas both GPON and Point-to-Point connections will ultimately be available.
I hope that this explains the architecture of the UFB network, and how connection bandwidth is dimensioned. It’s not necessarily a simple concept to grasp, but with the misinformation that exists I felt it was important to attempt to write something that can hopefully be understood by the average internet user. The varying plan options and pricing options means that end users have the option of choosing the most appropriate connection type to suit their needs, whether this be a high quality business plan with a high CIR, or a lower priced residential offering that will still deliver performance vastly superior to the ADSL2+ offerings most users have today.
And last but not least I have one thing to add before one or more troll(s) posts a comment saying fibre is a waste of time and complains about not getting it at their home for another 5 or 6 years. UFB is one of NZ’s largest ever infrastructure projects, and to quote the CFH website:
“The Government’s objective is to accelerate the roll-out of Ultra-Fast Broadband to 75 percent of New Zealanders over ten years, concentrating in the first six years on priority broadband users such as businesses, schools and health services, plus green field developments and certain tranches of residential areas (the UFB Objective).”
Residential is not the initial focus of the UFB rollout, and never has been. Good things take time.