Redundancy in the FTTH network, or not?

Submitted by fredrik.nyman on Thu, 01/31/2019 - 11:00

One of the bigger questions when deciding on the architecture of the active FTTH network is; should it be redundant?

On the one hand the fibre network is used to connect a lot of residential customers but individual residential users tend to not be considered "important" from a redundancy perspective. If a neighbourhood is lost due to a cut fibre that is often viewed as an acceptable loss in a city-wide network. For the affected end-users (particularly any teenagers in the area) the problem is enormous, but from the operator perspective the extra investment to avoid those days of outage often comes down to a business calculation where it is considered too expensive to built redundancy into the design.

Of course a redundant setup adds some cost. Dual uplinks means dual fibre connections, more optical modules and more ports in the aggregation layer to connect both ends of the redundant architecture. But over the life-span of the installation the investment costs are negligible.

Image removed.

More often the operational challenge of a redundant topology becomes the main argument. Adding the complexity of redundancy protocols such as spanning-tree leads to a higher risk of failures in the network - configuration errors or unanticipated side-effects of changes in the topology making the redundancy protocol itself cause outages.

There are two camps in on this topics. The star-topology camp that prefers to keep the network architecture clean and simple, believing that less complexity gives higher availability. When switches are installed in a fibre shed in the neighbourhood chances of failure of an individual switch or the connection (within that fibre shed) to that switch is infinitesimal and thus a redundant connection adds complexity, potential for loops, more configuration of VLANs and so on that increase the risk of failure more than it reduce the impact of failure if it happens.

The ring topology camp on the other hand prefers to keep a redundant architecture to minimize at all times the outage time experienced by users for a single failure. The benefit of high availability even down on access level outweighs the additional complexity and cost. The ring topology camp can support more demanding business services over the same infrastructure because it by definition provide the redundant paths required by such services. This is not only a question of Internet connections but also connectivity for surveillance cameras, monitoring of critical infrastructure (electricity, gas, cooling..), alarm systems etc.

One contributing factor is the fact that redundancy on layer 2 is more complicated to operate. Spanning-tree and the other redundancy protocols do their job but they have limitations that needs to be taken into account when building and changing the network. The complexity is  actually mostly related to the non-hierarchical address structure of Ethernet layer2, VLANs and how Ethernet works on a fundamental level. The redundancy protocols themselves are not the only culprit.

Many Waystream customers are in the ring topology camp and have made the decision to run their network on layer 3 as far down into the access as possible. Redundancy on layer 3 is easy. In fact it is encouraged and goes back to the myths about the origin of Internet - back to the US. Defence Advanced Research Projects Agency (DARPA) that supposedly was tasked to design a robust communications solution to withstand the impact of war on infrastructure. True or not, the Internet Protocol provides an address hierarchy and signalling mechanisms to make redundancy easy, stable and reliable. Billions of users experience no problems on a daily basis over the most interconnected network the world has ever had - the Internet, which is all run on layer 3.

This shows that the perspective on redundancy or not shifts if you go layer 3 instead of layer 2. What is complex, fault prone and risky on layer 2 is embraced, reliable and a given on layer 3.

Either way, our customers have the monitoring data to back up their decision that a redundant topology provides very high availability for their customers. Considering that the infrastructure we are building over fibre will carry all communication in the future and the implications that has;We will entrust our safety to fibre networks. We will depend on fibre networks for our income. The fibre network will be central to accessing the public services we expect and need from our society. Sensors, alarm system, building control, e-health…. This is the perspective to consider when answering the question redundant or not?

The network is not stronger than its weakest link. I think it is the right decision to build the network redundant all the way to the access port connecting the user.

Blog posts

How do you troubleshoot IoT devices?

Submitted by fredrik.nyman on Fri, 02/15/2019 - 13:00

Continuing on the subject of troubleshooting the network. Troubleshooting MPEG video has the benefit of a user that can tell you if it doesn't work and you can simply ask that user if the problem persists once you have fixed it. But what if there isn't any obvious way to determine if things are working, for example is that trashcan really signalling that its' full or does the temperature device really update the building climate control properly?

How to see what your users see

Submitted by fredrik.nyman on Mon, 02/11/2019 - 10:21

Live broadcast TV is one of the most popular services in fibre networks. You can get high quality pictures because there is enough bandwidth to send video uncompressed. But the nature of broadcast media is that it is very sensitive to packet loss or jitter. There is no retransmission of packets because it is live – you can’t hold the stream to get a lost packet back.

FTTH is not like any other network

Submitted by fredrik.nyman on Fri, 01/25/2019 - 13:34

If you are working in network engineering, hands-on with the routers and switches in the network, you probably have seen your fair share of network problems. However well you build it there is always some intermittent issue, some complaining user, some application that doesn’t get the throughput, some website that is unreachable.

It’s part of the everyday chaos of running a network to deal with big and small issues.

The Way Better Blog

Submitted by fredrik.nyman on Fri, 01/25/2019 - 10:02

In this blog I will be writing about some of the topics, big and small, facing network engineers and fibre networks and the kind of challenges I have encountered working with our customers over the past 20 years or so.