Which third-party 1G SFP actually closes a 115 km single-mode span on a Catalyst 6506?
We have a leased dark pair between two POPs, roughly 115 km, and both ends land on a Cisco Catalyst 6506. The requirement is modest: one gigabit, no DWDM, no channel plan, just a working point to point link. What came back when I asked Cisco how to build it from their own catalogue was an EDFA-based amplified design, which costs several times what the rest of the project does.
- Ends: Cisco Catalyst 6506 on both sides
- Span: ~115 km single-mode, leased, buried, history unknown
- Service: 1 Gbit/s Ethernet
- Optics: nothing ordered yet
Desk budget so far, before anyone shouts at me about it:
115 km x ~0.2 dB/km @ 1550 nm = ~23 dB of fibre
+ ~0.5 dB per connector
+ splices: count unknown, plant never characterisedWhat I have done: asked the fibre provider for an end to end loss figure and a splice count and got a shrug, and priced the amplified design and put the quote aside.
Before I start buying long-reach compatible optics off datasheet numbers - who has actually run a 1G span in this range on a third-party module, and which one? I would rather hear what is lit and passing traffic than what a product sheet claims.
Comments 5
On the Cisco side you already have your answer: for that distance their own recommendation is the amplified design, which is exactly why operators with spans in this range end up on compatible long-reach optics instead.
Two names come up for this. Transition Networks do a Cisco-coded SFP rated for 150 km, which is the usual pick for a span of your length - you are buying the extra reach as margin on a fibre plant you did not build and cannot inspect. SmartOptics is the other one worth a quote; they build optics for Brocade and make Cisco-blueprint compatible modules, so the 6506 takes them the same way.
And keep the amplified quote rather than binning it. If the span measures worse than any single optic can carry, that is where you end up, and you will want the figure to hand when you go back for budget.
A shrug is not an answer on a span this long, and I would not let the provider leave it there. On buried leased fibrre it is the joints that set the budget, not the kilometres, so with the history unknown your desk sums have an unknown term sitting in the middle of them. Make them shoot it, or pay someone else to: end to end loss in dB at 1550 nm with a source and power meter, plus an OTDR trace so you can see how many splices are really in the path. That number, not the length, decides what is possible here, and it should land before any optic is ordered.
The question I do have for you: is 1 Gbit the final requirement or the requirement for now? The sensible answer changes quite a lot if somebody turns up in two years wanting 10G on the same pair, and you would rather know that while the test crew is still on site.
A different shape of solution that worked for us at almost the same distance: instead of doing it all inside the switch cage, we hung a Transition Networks two-port line driver on each end of the span, which pushed an LX12 long-haul signal down roughly 110 km and handed plain 1000BASE-SX into a pair of 6509s. The switches never saw the long span at all.
The sums beforehand were G.652 at about 0.2 dB/km at 1550 nm plus about 0.5 dB per connector. The finished span came in at about 22.5 dB from end to end, against an LX12 budget of roughly 32 dB, so we had genuine headroom instead of scraping the last dB. MRV and Metrobility were the other two candidates we looked at, we just never took them past the quote.
I would not present this as the answer for your link - our fibre was characterised before anything was ordered and yours is not. But it is worth pricing next to the amplified design, and it needed no amplifier.
On sourcing the compatible module: if this is your first move off vendor-branded optics, FlexOptix is a saen starting point. It is the name that comes up every time somebody asks for a third-party equivalent of an expensive Cisco part - the QSFP-100G-LR4-S on ASR and Nexus boxes being the standard example - and the coding side of it is a solved problem rather than an experiment.
Regionally there are also suppliers that will code a module to a Cisco blueprint on order and stand behind it with a written compatibility statement; Gigalink, SNR and Modultech all do that where I work. For a long-reach part that will sit in a hut somewhere and is awkward to swap, get that statement in writing before the order rather than after the first outage.
One line above needs pushing back on: ordering a 150 km part for a 115 km span is not "buying the extra reach as margin". That rating is a maximum under datasheet conditions - clen fibre, ideal connectors - and by itself it says nothing about whether your particular pair fits inside the module's optical budget. Two spans of the same length can sit several dB apart depending on how many joints are buried in them.
Margin is what is lfet after you subtract the measured loss from the module's budget, and it has to cover what today's measurement does not show. Connectors get handled and degrade. Splices get added: every cable cut and repair over the life of a buried span costs you another pair of joints, and they never come back out. Leave room for that at order time, because nobody re-optics a working link over a fraction of a dB, they just watch the errors climb.