Keeping product lines around

Submitted by fredrik.nyman on Fri, 03/15/2019 - 09:50

Building fibre to the home networks are different from any traditional enterprise or telecommunications network. One of the main differences is the time it takes to complete the network. You make a plan, design a an architecture with VLANs and redundancy and imagine how this will scale as the number of connected customers increase. But then the years go by, because building a fibre network to connect every home in the community can take decades.

For example, one of our customers in Sweden started their deployment at the turn of the century. They are now close to 100% connected homes in their region, but it has taken almost 20 years to get there.

Obviously things change over that time horizon. New models of switches, more powerful with new capabilities. Services evolve, customers use more bandwidth and there are more devices to connect now than at the outset.

Image removed.

One of the particular challenging aspects of the network is that you will go through multiple generations of access equipment as the network is built out. Not least for capacity reasons there will be generation shifts in hardware over such a time period. While it may not  be necessary because of user demand, vendors rotate their product portfolio and new models are of course better for the most part but it is not so good if it means big changes to configuration, architecture, number of ports and so on because it might have more dramatic impact to the network operation than first glance gives.

As one of our customers put it when discussing introducing a 48-port model to complement the current 24-port switches, I thought they would appreciate the higher density - fewer boxes to install on remote sites but they highlighted a concern. They have fibre based installations and they have copper based installations. Their field engineers have one copper and one fibre switch as spare-part as part of their "kit" in their vans. Since the customer covers a large region the field guys have their vans with them home when on duty so they can drive directly to the location if there is some issue. Introducing a 48-port model when their entire network today have 24-port units would mean that all their field guys would have to squeeze another model into their "kit" because they don't know beforehand what type of unit they might need to replace when they go out on a call. Driving first to the central warehouse to pick up the right box adds time to fault resolution which is not acceptable. The logistical concern was certainly valid.

The same thing can be observed as vendors introduce new models of their switches to replace older products. It is not uncommon that some vendors release new models that replace their current line as often as every 3-4 years. If your network takes 20 years to build you will might end up with 5-6 generations of products in the network. Add to that if you have a multi-vendor strategy and the vendors change configuration syntax, have different firmware images, stops supporting units after end of sale and make physical changes to port layout and you have a recipe for a reality many network engineers face.

Chaos.

Feature incompatibility. Reliability issues because different switches behave and work differently.

Yes, the new model have much better features (that you probably won't use because you can't roll out functions in only parts of the network particularly if it would mean that certain services only are available in the neighbourhoods that got connected to the new generation) but if it means moving the uplinks from left to right and power from the rear to the front you might need to rebuild every old site when replacing with a new unit - new fibre patches because the old ones were too short and new electricity wiring.

 

At Waystream we know that our customers have long deployment cycles. Therefore we try to maintain our models for an extended time - keeping physical form-factor, appearance and features compatible even if hardware components within the products change. This gives apparent long life-spans of access products, something that fits very well fit our FTTx customer base - they can work with the same product line for many years.

Of course generation shifts in Ethernet speeds ultimately force big changes, but we keep our model numbers around for longer than most other vendors. It makes it easy for purchasing to buy the same make and model. It makes it easy for logistics to deal with well-known products, maintain stock levels to cover new builds and keep spare-parts available.

Sure we are doing changes to adapt for component updates or increase performance, but we keep them backwards compatible and keep the physical characteristics (ports, power, cooling, dimensions) to let all the non-network techies (purchasing, field engineering, logistics, project management, bosses, finance…) feel familiarity.

Backwards compatibility is also important for the networking team. Automated deployment, configuration copy/paste, control/provisioning systems, network management systems… all those different support systems and tools works better when they don't have to be updated and readapted for a new generation access switch every so often.

The longer you can run the units in the network the better your business case becomes. That includes avoiding too many changes and updates to support system, retraining staff, lost time and confusion when new models are introduced or somebody brought the wrong box to the new installation.

Yes you can upgrade a 7 year old Waystream switch to the latest version of our iBOS firmware and gain the new features also on that unit. At least as far as the technology reasonably can support it.

We're now moving towards our fourth generation as bandwidth demand continues to grow, but we still have customers happily running on our first generation switches designed almost 20 years ago. That is reliability.

Blog posts

Detecting service quality issues

Submitted by fredrik.nyman on Tue, 03/24/2020 - 15:31

While networks handles the increased load from Corona with ease, not all services can say the same. Detecting quality issues in service delivery is much more difficult than checking if the network itself can handle the load. There is probably a million reasons why service quality can degrade even if the network seems to be working fine. The trick to detecting and troubleshooting these situations is use of telemetry and service assurance.

European networks unvoluntary casulties of US-China trade-war

Submitted by fredrik.nyman on Mon, 09/02/2019 - 13:19

I usually don't comment on our competition but recent events such as new legislation proposal from the government in Sweden has sparked a well-needed debate. I commented on the situation for Swedish city networks last week in this IDG article https://computersweden.idg.se/2.2683/1.722350/konkurrent-varnar-huawei-stadsnat and also wrote a debat

Turn on automation of your FTTH network

Submitted by fredrik.nyman on Mon, 04/01/2019 - 09:08

The distributed nature of a fiber to the home network means that you will have equipment spread out and you might not always do the on-site installation yourself. If every switch has to pass your desk for pre-configuration port before getting deployed into the field you will need to deal with the logistics of getting the units from your warehouse via your desk, packing and unpacking, and clearly marking them so that the right unit goes into the right location.

SDN and NFV in FTTH

Submitted by fredrik.nyman on Thu, 03/21/2019 - 09:51

I love acronyms. You got three of them in the title of this post.

In recent years we got Software Defined Networking (SDN) and Network Function Virtualization (NFV). Many of the large telcos have invested millions into research of these subjects and are pushing the industry in this direction. Telefonica has expressed high ambitions to move to a completely SDN/NFV enabled network in record time. All the big ones are involved.

Keeping product lines around

Submitted by fredrik.nyman on Fri, 03/15/2019 - 09:50

Building fibre to the home networks are different from any traditional enterprise or telecommunications network. One of the main differences is the time it takes to complete the network. You make a plan, design a an architecture with VLANs and redundancy and imagine how this will scale as the number of connected customers increase. But then the years go by, because building a fibre network to connect every home in the community can take decades.

Save the planet - work from home

Submitted by fredrik.nyman on Thu, 03/07/2019 - 10:30

In my last post i revealed how dirty a fiber network can be depending on the source of electricity powering the network. I showed how a typcial 24-port access switch might contribute anything between 23kg to 485kg of carbon dioxide per year to the atmosphere depending on the electricity mix and how that can be reduced with lowpower optical modules.