Single fiber between buildings: how to link a 1G SFP on an Avaya VSP 4400 to an SFP+ port on an Extreme X690
Joining two campus buildings into one network. In building A there's an Avaya VSP 4400, and only 1G SFP ports are free there: the dedicated uplinks have long been taken for other purposes. In building B there's an Extreme Summit X690-48x-2q-4c, and there only 10G SFP+ ports are free, no other options either. There's a single OS2 fiber laid between the buildings, no way to get a second strand.
- Avaya VSP 4400, free 1G SFP ports
- Extreme Summit X690-48x-2q-4c, free 10G SFP+ ports
- run: single-mode OS2 fiber, one strand between buildings
- modules for this line not yet purchased
Measurements along the run and current port state:
Run: single-mode OS2 fiber, one strand between buildings
Attenuation at 1310 nm: about 1 dB
X690: SFP+ 10G port, link doesn't come up with a 1G module
VSP 4400: SFP 1G port, freeWhat I've already done:
- went through both chassis looking for free dedicated uplinks - none exist, so "plug into the right port" isn't an option
- on a bench, put a plain 1G SFP into the SFP+ cage on the X690 and connected it to the VSP 4400 with a short patch cord: the port just stays down
- dug into the X690 datasheet's 1G optics tables but never fully worked out the single-fiber scheme
Two-part question: which modules should I get for a single strand so a 1G module in the Avaya matches up with the SFP+ port on the Extreme, and what still needs configuring on the X690 side?
Show original (Русский)
Also available in: Русский
العربية
Deutsch
English
Español
Français
हिन्दी
Bahasa Indonesia
Italiano
日本語
한국어
Nederlands
Polski
Português
Türkçe
Tiếng Việt
简体中文
Comments 4
Since there's only one strand, plain modules are out - you need a matched pair of single-fiber BiDi modules. That's two modules with mirrored wavelengths: one transmits on one wavelength and receives on the other, the second is the reverse. On their own they're useless, buy them strictly as a matched pair.
Second, and more important: pick the reach class for the actual run, not with margin for the future. At about 1 dB of attenuation, a long-reach module can easily blast out the receiver on the other end, and instead of a link you'll get errors or the same down state. If after installation the receive level turns out above the module's datasheet upper limit, put in an attenuator - on short runs between neighboring buildings that's standard practice, not a workaround.
On configuring the X690: don't expect the SFP+ cage to negotiate down to gigabit on its own. You need to manually lock the port speed to 1 Gbit/s, and then the 1G module works properly in it. Check the X690 datasheet for exactly which 1G modules the cages accept - there are separate tables for 1G optics there. Both chassis, by the way, work fine with non-original modules, so there's no point insisting on vendor transceivers for a single line.
Clarify a couple of things, otherwise it's guesswork. First: how was the attenuation measured - a light source with a power meter, or an OTDR? About 1 dB on single-mode is very little, and whether you can trust that figure when picking a reach class depends on the measurement method. Second: on the X690 did you leave the port on auto, or try forcing the speed? The SFP+ cage sits at ten by default and doesn't always come down on its own. And third - what's the run length between the buildings? On neighboring buildings a long-reach module can easily blast out the receiver on the other end.
Answering the question above right away: attenuation was measured with a light source and power meter, no OTDR on hand, so the figure is just the loss from patch panel to patch panel.
Got a pair of single-fiber modules with mirrored wavelengths, on the X690 manually pinned the port speed to 1 Gbit/s - the link came up on the first try, no fuss. Receive levels are within spec, no attenuator needed, the run really is short. Been running traffic between the buildings for a second week now, the port doesn't blink, no loss.
I'll add something about the Avaya side, so nobody gets a surprise later when carrying the config over to a neighboring chassis. The autonegotiation rule on VSP differs from model to model. On the VSP 7254XSQ there's simply no autonegotiation on these ports: a 1G SFP there links up only if auto is turned off on the far end, and copper 1000BASE-T modules on this platform work strictly at gigabit. On the VSP 4450 Series it's exactly the opposite - for a 1G SFP to come up toward third-party switches, autonegotiation has to be left on. You can't stretch one configuration template over both platforms, so check the rule for your specific model first, and only then blame the module or the port.
And yes, locking the speed isn't just an Extreme thing. On a Nexus 3000, a copper 1G SFP in an SFP+ port simply returns
Ethernet1/45 is down (SFP validation failed)until you go into the interface and setspeed 1000andno shut: there's no automatic 1G detection there, the port defaults to ten and rejects the module.