How to choose a transport(terrestrial/subsea)

Thank you everyone incredible amounts of responses for my how to choose a transit provider smail earlier.

How do you choose transport & backbone?

Looking at key aspects like route information, diversity, aerial vs under ground fiber, age of fiber, outage history, length, but what else?

I will get both transport and transit as two seperate blogs.

I will also submit as a nanog paper for the meeting after next, or maybe next? I am probably too late by now.

Thank you for all your help. I will add your names to the thank you line :wink:

I also look at hand-off locations (as long as it doesn’t compromise the overall robustness of the design).

Most providers will be able to hand-off in the BMMR of a carrier hotel and some will have the flexibility to hand-off in particular suites within the same building or other locations near where the cross-connects fees are lower. I’ve seen cross-connect fees between $50 up to $750 MRC so if you need multiple wavelengths (for capacity), the cross-connect fees are going to make a huge difference on the total MRC.


heh, cross connects are indeed a major issue. I have a need for > 10G transport. My equipment supports 40G. The carriers aren’t terribly interested in doing 40G transport (at least not at a reasonable price, one quote was over 4x a 10G). 100G-capable switches cost too much. Equinix charges as much for a pair of cross connects as a 10G wave. Carriers aren’t likely to be interested in using bidi optics or passive WDM to overcome the ridiculous cross connect charges.

This all complicates how one chooses transport. There’s no easy path forward.

I think the Juniper MX204 is not a bad way to deliver reasonably inexpensive 100Gbps ports to customers. The box is reasonably priced and is, essentially, an MPC7 in a pizza jacket. If an operator is not overly religious about what box they hook high-capacity customers (40Gbps+) into, the MX204 is a good way to start offering affordable 40Gbps and 100Gbps services. Mark.

No cost affective 10x10G to 100G muxponder?

Luke Guillory
Vice President – Technology and Innovation

Tel: 985.536.1212
Fax: 985.536.0300

Reserve Telecommunications
100 RTC Dr
Reserve, LA 70084

The information transmitted, including attachments, is intended only for the person(s) or entity to which it is addressed and may contain confidential and/or privileged material which should not disseminate, distribute or be copied. Please notify Luke Guillory immediately by e-mail if you have received this e-mail by mistake and delete this e-mail from your system. E-mail transmission cannot be guaranteed to be secure or error-free as information could be intercepted, corrupted, lost, destroyed, arrive late or incomplete, or contain viruses. Luke Guillory therefore does not accept liability for any errors or omissions in the contents of this message, which arise as a result of e-mail transmission.

FS had one, but it’s not on their site anymore.

Mike have you looked at Packetlight? Long-haul is mostly jumping to 100 or even 400g coherent.


I haven’t.

Sure, but the equipment still does smaller channels. Going to 100G or 400G for just over 10G seems silly.

If Equinix had reasonable cross connects, I’d just LAG 10Gs. The cost of a pair of Equinix cross connects isn’t much less than the 10G wave. Thankfully I’m only in one datacenter with such a ridiculous model.

Back to main discussion…

How do we choose the best transport?

One question, how much people care about vendor diversity? I do and did care. I don’t want to put all my eggs in one basket. Do you care? Thank you


As long as you understand that vendor diversity doesn’t imply route diversity. Diversity within a given vendor is still subject to the same chassis, the same automation platform, the same billing department, etc.

There are advantages and disadvantages to vendor diversity.

As advantages, you won't be subject to complete loss of connection because of a single dispute or provisioning/control plane issue with that one vendor. You can also more easily pit vendors against each other for pricing if you are already vendor-diverse.

As a disadvantage, not only does vendor diversity obviously not imply route diversity, but it will completely put the onus on you to ensure route diversity if you want it. With a single vendor, you can demand that your circuits have route diversity and, assuming you trust them, they have all the information they need to make that happen for you.

I can’t stress enough the importance of controlling your own route and even cable diversity. Require KMZs of the routes for any services you take (especially single path Wave type services). Put them in the contracts if you can.

I’ve had at least 1 situation where we had vendor diversity and what was supposed to be route diversity- 3 separate waves coming south and southeast out of a datacenter to 3 separate cities. Imagine my surprise when we took a outage one day that severed all 3 circuits. Yes all 3 circuits, going to 3 separate cities, on 3 separate carrier/s DWDM platforms, all happened to show up in the same sheath of cable at one location that happened to experience backhoe fade. Was not a good day…

That’s a great example. Thank you James for sharing. I have done so many “GROUND TRUTH” visits where randomly selected certain physical points to validate physical diversity. Have seen several places where dual risers in the building were present or multiple building entries were available but not used. Ground truth events are certainly important and can be eye opening. It does not necessarily scale as you can’t really walk all the fiber A-Z everywhere… i know.

If people start spot-checking this stuff more regularly, perhaps the companies being verified will take delivering the correct product the first time more seriously.

Some of it boils down to a lack of data quality about what they actually have.

The Trans-Atlantic cables, particularly on the UK side, probably lack physical diversity. The landings at Bude probably share back haul. The cost of each cable digging its own trench was quite high.

  • R.

Some of it is due to lazy buyers purchasing two IP ports from distinct companies without considring that two ports both located at the site are vulnerable to shared risers or entrance facilities.

  • R.

More likely everyone bought IRUs out of the same ILEC’s single cable.

Or they just all go through the same single raceway to enter the building, etc.


  • Ben Cannon, AS15206

I guess today shows how important vendor diversity can be. :slight_smile:

I’m of the opinion that, if you need resiliency, you should order explicitly diverse circuits from a primary provider and then a secondary circuit from a second vendor.

Ultimately, If you want contractually-enforced physical diversity then the best options will be single-vendor solutions: Obviously you also want to avoid an unknown single-vendor single-point-of-failure, hence the secondary provider. Having two vendors is usually a less than optimal solution since neither has visibility into the others’ network to ensure the physical diversity required for a truly resilient service: what happens if an undersea cable is cut, etc?

The cost of such solutions is often unpleasant to justify, mind.


All true but it is becoming increasingly difficult to determine if a provider is using another providers infrastructure (all are at some level). For example, in the SIP world there are several national level carriers that are using Level 3s core SIP network and if you were not aware of that you could buy trunks from two of the largest SIP trunk providers in the US and actually be running on the same network. Carriers are also very often reliant on the ILEC for fiber and last mile access. Especially in non-metro areas getting diverse last mile access could be impossible or have huge construction costs. It is pretty complicated to ensure that your carriers are really diverse and much harder to ensure that they stay that way. I have many examples of carrier grooming their own primary and backup circuits onto the same L1 path and not realize they have done so.

Contractual diversity is a great idea that does not work since the carriers do not actually know what each other’s network looks like. So let’s say that Sprint and CenturyLink choose the same fiber carrier between areas, do you think they would notify each other of that fact? Do you think the fiber carrier would tell them what another customer’s network looks like? You can tell Sprint to not use CenturyLink but there is no way to get both of them not to use the same third party. I suppose you could contractually tell a carrier to avoid xxx cable but I would have little faith that they maintain that over time. I seriously doubt they review all existing contracts when re-grooming their networks.

Steven Naslund

Chicago IL