In theory I like the SFP concept very much. However what strikes me is the manufacturer restrictions for every switch/card. How come different SFP modules work with different vendors only? Is there a design flaw in the SFP specification? What makes the SFP modules vendor dependent?

Side question: Are vendors being mostly compatible with each other, or are there vendors acting evil in the sense of being known to be trying to restrict their compatibility?

3 Answers 3


How come different SFP modules work with different vendors only?

It's actually the other way around. SFP modules have no logic to see where they are used. Network devices may check compatibility and frequently do so, with many vendors deliberately overdoing it.

SFP modules contain a small serial EEPROM that stores a standardized structure to describe the module - (fiber) type, wavelengths, reach, vendor, SKU, serial number, manufacturing date, ... There's additional room for vendor-specific data. The structure is standardized in INF-8074i clause B.4 (slightly modified from the earlier GBIC standard and common to a wide range of interface module types).

That data is commonly used to accept or reject generic modules in many devices. This isn't a design flaw, just a decision that a device vendor takes for compatibility and monetary(!) reasons. Technically, there's little point in doing so for standard modules. There's only a handful of actual optical module manufacturers anyway, nearly all vendors use OEM.

Some devices simply don't care what you insert and accept anything. Others can be configured to ignore the SFP type/vendor, reducing your vendor support. Yet others reject 'alien' modules in general, forcing you to buy compatible ones - either official ones by the vendor (sometimes very pricey) or third-party modules flagged as compatible. There are even module vendors that let you field configure the SFPs you buy from them - very handy if you don't like stocking a module zoo.

Unfortunately, we cannot recommend any vendor here due to site policy.

Personally, I have very often chosen to buy multiple compatible modules (for spares or redundancy) instead of a single 'original' one. In a project, the price difference may become negligible, but bought seperately I've seen the 'original' part being marked up more than fifty times the price of a quality 'compatible' part.

  • 3
    There are also devices which do the exact opposite: they don't do any checks at all, and will happily accept e.g. a DAC even though they only support fiber. Jan 19, 2022 at 22:27
  • Talk to your sales rep--there might be a back door command to allow 3rd party SFPs. Be aware that there are cheap "bootleg" SFPs, and SFP do commonly fail.
    – Andrew
    Aug 12, 2022 at 10:16

In reality they are very compatible.

I almost never use the official vendor SFPs and never had any issue (but I always use two well known third-party SFP vendors).

There's two main points for vendors to promote the use of their own SFPs:

  • to guarantee the solution will work
  • pure marketing / money

The first point is mainly to avoid any potential issue, cause they cannot offer any guarantee that a third party product works as expected with their switch.

If you have a support contract for you switch and use a non-vendor SFP, if you contact the support they will likely always blame the third-party SFP first.

The second point is obvious enough I think.


Vendor lock-in is a matter of greed / marketing / etc. No system vendor actually makes optics. They may have tested a specific model thousands of years ago. (With the exception of maybe HP/HPE as they own(ed) Agilent.) Most don't even see the optics on the way to the customer. They're labeled Cisco, Juniper, etc. at the manufacturer and sent to the various wholesale warehouses. That $1000 optic you "bought from Cisco"? Did not come from Cisco. Did not come from a Cisco warehouse. Did not even fly over a Cisco office. (probably.)

How they are vendor-locked (i.e. "compatible with X") is via a signature stored in the user area of the SFP EEPROM. (The Cisco method is publicly reverse engineered. And has 64 vendor keys!)

[My former employer did the same thing. They paid Finisar and Methode to put the company name on them - no signature. In their case, there were trays of them within the company's stores. I had two trays of them in my office.]

  • It's not always about greed. I have some HP switches that were originally open. However, the original HP-branded SFPs had a serious issue and were discontinued, replaced with a "b"-series model. A firmware update for the switches then began checking to make sure SFPs were marked as "b"-series or later (there's also a "c" series, IIRC). So what was open originally later effectively become vendor-locked. I can (and do) still use generic SFPs, but I have to watch for generic models that are aware of the issue and mark the device correctly. Jan 20, 2022 at 14:52
  • 1
    Nope, that's still a greed inspired answer. Instead of blacklisting a known bad module, they opt for the industry wide practice of "must buy our super-overpriced blessed versions".
    – Ricky
    Jan 20, 2022 at 15:24
  • 1
    @JoelCoehoorn, do you mean the J4859x SFP LX transceivers (and others)? They at least have those three versions, A, B and C, and I also remember older HP switches accepted ~any SFP, while newer ones required the B version. Never saw any real issues with any of them in use though, so it appeared just pure lock-in greed... (The A ones didn't seem to support monitoring though, but they still work.)
    – ilkkachu
    Jan 20, 2022 at 16:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.