I have a lot of special needs clients (should tell you something right there). Many of our customers have equipment colocated with us and want to use the wireless as a sort of backup point to point connection to their racks. Not a big deal, we VLAN the traffic and run a cable to our NOC, life is good. For whatever reason, we keep getting requests for this service and running cable down to the NOC. My thought is to run fiber from the radio room to the NOC and split it off in the NOC instead of the radio room. I guess I'm getting tired of running Ethernet cable and this is not something we envisioned when we started. But more and more customers want backup links to their equipment or backups for their internet and want to run BGP. Anyway I would like to use Mikrotik and I would rather not introduce a point of failure by using fiber converters. So my question is, is there a fiber Mini PCI that is compatible with Mikrotik? Another option? If I have to use converters than I have to use converters. Any suggestions would be appreciated.

I guess the only way to mate Mikrotik and Fibers right now is using x86 equipment and intel network cards with SFP slots on them.
There is no such thing as a mini-pci SFP, to be honest the PCI bus would be way too slow for Gigabit Ethernet anyways.

Sooo.... just put up with one of those so-called "carrier-grade" fiber converters, otherwise be prepared to fit a 400$ network card into an x86-based machine
PS: Even the SFP itself can be seen as single point of failure... and SFPs (from my experience) fail way more often than really good fiber converters.
Still, the SFP is a more elegant solution to the problem

hedele,
Thanks for the response! I had done some research and I thought that was the way I was going to have to go, but thought, i might get luckey and be wrong on my assumption.

There is still the remote chance of Mikrotik releasing an updated Routerboard 1000/1100 with one or two SFP slots.
At least there has been a lot of begging in the Forums and MT seemed to consider thinking about building something like that

currently we have a lot of devices coming out soon, once that has calmed down, we will thing about it. we have to finish the started projects.

Thanks Normis,
Is there any chance of wetting our appetite with some insight about the new products?

I tend to avoid media converters. I've used a number of different brands before and I still have no clue how they can all make solid state devices that fail so often...

MCT,
They do seem to be strange little beasts; they either work forever or fail quickly.

MCT,
They do seem to be strange little beasts; they either work forever or fail quickly.

I agree with that. In fact that does apply to almost any piece of hardware..

I tend to avoid media converters. I've used a number of different brands before and I still have no clue how they can all make solid state devices that fail so often...

The predicted mean time between failure (MTBF) for Laser Diode Modules can be as low as 5000 hours. Todays fiber optic converters should have an MTBF between 50,000 to 100,000 hours. It's not unusual to see MTBF of 1 000 000 or more. Cisco for example, the predicted MTBF on their SFP modules is 12,820,512 hours. I'm a little bit sceptical on that number, but it's what the data sheet says.

http://www.cisco.com/en/US/prod/collate ... ba88e.html


I would suggest that MikroTik base their products on hot-pluggable I/O devices that plugs into module sockets, so the users can decide the type of modules to use, and replace them if they wear out.

I think this sounds a little bit fishy, 1500 years MTBF, and only 90 days warranty.

I think it's 12,820,512 seconds, not hours. 24 years sounds more realistic.

currently we have a lot of devices coming out soon, once that has calmed down, we will thing about it. we have to finish the started projects.

Remember to use slots like the RB800 ones to the minipcis of boards like 433s please!

I think this sounds a little bit fishy, 1500 years MTBF, and only 90 days warranty.

I think it's 12,820,512 seconds, not hours. 24 years sounds more realistic.

That's not how MTBF works. http://en.wikipedia.org/wiki/Mean_time_between_failures

MTBF is a slightly useful statistical failure model and has nothing to do with predicted life span.

If you have a 50,000 hour MTBF, and you have 1000 optical modules (say you have a large university campus network), you get 50,000 / 1000 = 50 hours. On average you will get one failure per 50 hours.