0

We are looking for upgrade for our small 10G network, and for that we need to hook up computer via optical link to 8-port 10G Netgear XS508M switch (it's mostly copper RJ45 switch, but one port is RJ45/SFP+). Plan is to pop in PCIe card (Intel X520-DA1) into computer, get about 10 meters of MMF optical cable, and get two transceivers, SFP+ LC SR (short-range) because we need to cover very short distance.

I am looking at prices and transceivers (same LC-MMF SR) have wildly different costs on Amazon. For example, Netgear's one is $300, Cisco sells same for $1200 or just $70, when what people call "off-brand" goes for $22

I wonder what is the difference and is there incentive for buying more expensive option? We did learn that cheap person pays twice, and use server-grade hardware as much as possible, however, for since transceivers have no moving parts, is it safe to pay up to 10 times less? Especially with combo of Netgear 10G switch (it seems that HP switches are picky)

PS: this text describes (rants?) price difference but only mentions that "higher OEM parts' price comes with affirmation of testing for compatibility" without much more details.

1

There are several sources for transceivers:

  • branded, "original": supplied by the device (switch, router, ...) vendor, these are guaranteed to work in your equipment at all times
  • "OEM" / compatible: These modules are programmed to look undistinguishable to the original transceivers and will also work well. However, there's a small chance that a firmware update will change the way the transceivers are checked.
  • generic: These modules only work in devices that don't care about what you plug in. Sometimes, the branding check can be explicitly deactivated. All this can also change with firmware updates, however.

In addition, there are varying qualities for the transceivers. While there are only a handful manufacturers, they might select quality by testing ("binning") and sell off lower quality at a lower price. But we don't know this for sure. In practice, the quality range is rather small.

Interestingly, the branded, "original" transceivers are not manufactured by the device manufacturer but OEM and they are not necessarily of the best available quality. If you find a good source of compatible transceivers these are usually of at least the same quality.

This leaves you with the choice between reliability (including firmware updates), quality (durability, achievable error rate, reach) and price.

Well, we had the choice of buying original LX transceivers at 200€ or (good) compatible ones at 29€. Additionally, the vendor offered us compatible BX, single-strand SFPs for just 45€ while the original ones ranged at 700€. We ended up buying twice the number of compatible BXs as LXs planned and aggregated the links - cheaper, faster, and more reliable than the original LX due to redundancy.

  • are you saying that some transceivers are not passing (or miss by a bit) manufacturer's testing and being sold as "generics"? Sort of like 99% percentile tested are sold as Cisco for $500 and 95-99% percentile tested are sold as XYZ Cisco-Compatible for $20? – aaaaaa Oct 29 '17 at 22:04
  • 1
    I'm not saying they don't pass - I'm pretty sure all tested and sold modules meet the standards. Higher quality transceivers may have a bit more margin though (if you do need it). What these higher-or-lower quality modules are sold for, only the manufacturer and - maybe - the vendor know. – Zac67 Oct 29 '17 at 22:10
1

Basically what it gets down to for any new Infrastructure equipment regardless of brand/cost:

  • Buy evaluation unit(s)
  • Test
  • Evaluate results and take action.

Go ahead and buy a couple and see how they go. For $22/ea, it is a cheap as chips test. If they don't pass testing (and by testing I mean stress them with traffic so you get more than just a simple link test), you're only out the time. I also recommend testing the branded ones as well in a bake off with a couple of OEM vendors and the branded ones. Then keep the cheapest ones that pass, and return the other test/eval units. Always TEST! :)

I've got a regular/reliable supplier at this point for OEM transceivers where I'm shelling out $16-22 for transceivers and they're doing fine in several varied bits of Netgear, Miktrotik, Quanta, Brocade, Arista, Cisco, etc etc etc, in both work environment (where there is normally a reasonable budget) and in my home lab (limited budget). They've even worked well with Brocade Switches and CNA's which are insanely finicky. (I'd post the link, but that gets into making a product recommendation which is verboten).

  • i agree with "buy cheap and test" strategy. That is definitely the plan (and by testing i mean something like running traffic simultaneously both ways for hours). what I wonder as well is what makes some switches "finicky" and incompatible with cheaper transceivers. Is it that some switches have less tolerance? – aaaaaa Oct 29 '17 at 22:00
  • 1
    Most of the time, "finickiness" has to do with how strict the checks are for the uuid and other mfg info burned on the rom. It doesn't have as much to do with real world reliability. In Brocade's case, they actually lightened up for active twinax cable with more recent firmware updates on the cna's right before selling the division to Qlogic. Then you get switches like Netgear or NICs like Mellanox that will take anything. – Jeff Burns Oct 30 '17 at 3:27
0

As far as I can see, many off-brand SFP+ transceivers are second-hand. The temperature of new optical transceivers is usually at 0-70 ° C but many second-hand optical transceivers are inaccessible. And the second-hand optical transceivers cannot operate normally in high-temperature or low-temperature conditions.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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