Use Our Optics, or Else!

Cisco Systems Inc. (Nasdaq: CSCO) and Extreme Networks Inc. (Nasdaq: EXTR) have discovered a new revenue source that's got some customers steamed, and a grass-roots effort is attempting to intercept the trend before it spreads to other equipment vendors, Light Reading has learned.

It has to do with optical modules, the doodads that connect a line card to an optical fiber. These are frequently purchased from the equipment vendor, and they conform to multiservice agreements (MSAs) that dictate size, shape, and function. In other words, they're supposed to be interchangeable.

Sometime in the past year, Cisco equipment started rejecting Gigabit Ethernet small-form pluggable (SFP) modules that came from other vendors, say sources. This forces the customer to buy SFPs from Cisco at a hefty markup. Extreme reportedly has followed suit, and other vendors reportedly are considering similar policies.

Cisco and Extreme failed to respond to requests for comment. But Cisco has stated in company literature that the new policy was meant to protect customers. The company refers to its SFP-blocking feature as the "Cisco quality ID," according to a policy document obtained by Light Reading. The document says the quality ID "protects customers against mistaken use of non-qualified GBICs and SFPs (including counterfeit products) and ensures a fully qualified network configuration."

How could this pose problems? Consider the tale of one Cisco customer who has asked to remain unnamed. Last spring, the customer was replacing another vendor's equipment with Cisco's and planned to reuse the SFPs from the old equipment -- only to be told that Cisco's IOS software would disable every port that used the old SFPs.

The only alternative was to buy SFPs from Cisco directly. The tab: more than $300,000.

Shaw Communications Inc. likewise encountered the issue earlier this year. "We always were mixing and matching GBICs [the predecessors to SFPs] with customers' equipment, because it worked," says Dave Wodelet, chief network architect. Wodelet would not specifically name the vendor but described it as a major public Ethernet switch vendor. One day, the vendor's box started reporting errors for every foreign SFP. "When we approached the vendor, they said they'd had trouble with third-party SFPs, and customers asked them to do this."

Network operators like to swap SFPs around because it saves money. That's a long-held benefit of standards. Different SFPs handle different distances, for example, so a port can be upgraded to a longer reach by adding a new SFP.

The new policy seems to have snuck up on customers. According to sources, it only affects newer products.

Wodelet decided to see if other carriers were as irked as he was. He presented the issue at last October's North American Network Operators' Group (Nanog) meeting, and he's been invited to give a similar talk at the Optical Fiber Communication Conference (OFC) in March.

"What this has done is put a dividing line in the sand, and the companies who were thinking of doing it are waiting to see what happens," Wodelet says. "I'm sure if Cisco and Extreme got away with it, others would follow suit."

What really annoys Wodelet is that the equipment vendors don't even make SFPs. They all buy from the same small pool of suppliers -- Finisar Corp. (Nasdaq: FNSR) and Agilent Technologies Inc. (NYSE: A), mostly -- then store their own IDs and serial numbers on a memory chip included on the SFP.

This means that a rejected SFP could have come from the same manufacturer as a "Cisco" SFP; the only way Cisco knows the difference is by looking at the serial number. Wodelet says he's verified with Finisar that they send the same SFPs to everybody. In other words, a switch that rejects certain SFPs is like a flashlight that works with only one brand of batteries.

This type of practice isn't new. It's a common technique in consumer electronics, where manufacturers will use proprietary connectors, forcing customers to buy expensive replacements if, say, the power adapter to a portable CD player breaks (not that anyone's bitter about this).

Cisco hasn't turned a deaf ear to the complaints. "They said there has been a commitment to make a decision in January. They'll decide once and for all whether they'll stand by the current policy," says the Cisco customer.

This source estimates it would take at least six months to update IOS should Cisco rescind the policy, so it's not likely to change quickly. Meanwhile, Wodelet will continue speaking about it, trying to make users aware of the complications they might face.

Could Cisco sales be hurt by the quality ID flap? Not necessarily, but its pride and profits could get bruised. Upon learning that the Cisco equipment would require new SFPs, the unnamed source demanded they be thrown in for free, and Cisco's sales team -- which had already made concessions just to win the upgrade -- agreed. Problem solved. For now.

"As we expand our network and need more connections, we'll have to buy Cisco optics like everybody else," the source says.

— Craig Matsumoto, Senior Editor, Light Reading

Archives of Related Light Reading Webinars:

Page 1 / 4   >   >>
av8T0r 12/5/2012 | 2:44:02 AM
re: Use Our Optics, or Else! I just saw this and haven't read all the messages so I don't know if this was covered already. The MSA is SUPPOSED to guarantee that all SFP's work the same and interoperate with all hardware. Anyone with experience in our industry should know that's a pipe dream.

The truth is that there are numerous interoperability issues with SFPs not only between manufacturers but also between different revs from the same manufacturer.

Say a user buys a box from X company but wants to use SFP's he bought from someone else, on ebay , or where ever. What happens if the SFP's don't link? He'll call X company and want them to fix the problem. X company will probably do what ever it takes to solve the problem for SFP's they sold because they are guaranteeing they will work. How are they supposed to fix problems for parts they didn't sell and didn't test on their box?

The high margins OEMs receive in part cover the very high cost of supporting the SFPs, which, BTW, is a real pain (yes they want to make some money too - if they didn't the products wouldn't be offered). The SFP makers? From what I see, they are not well equipped to support end users, especially if they did not sell the part directly to the user.

I relate to end-user frustration on this issue. The reality is that SFPs are not totally interchangeable just like computer peripherals are not totally interchangeable.
Ben Crosby 12/4/2012 | 11:12:06 PM
re: Use Our Optics, or Else! It's a good article, but misses one key aspect altogether. Cost. I know, this can be religious, but...

The fact that few vendors produce these SFP's and the network equipment manafacturers OEM them is all fine and dandy, but obviously in many cases it is the volume purchasing of these modules by the network equipment manafacturers that allows them to design the more expensive plugability onto line cards than fixed optics.

That's to say that a vendor like the big C, buying tens of thousands of SFP's to re-brand and sell can negotiate a far better price than the RBOC/ILEC or whoever needs to use the modules. Looking at some pricing the discount to the vendor may be on par with their markup to their customer. i.e. C buys from F at list-% and sells to RBOC at (list-%)+% (i.e. - C's markup doesn't really affect the price that the RBOC would buy at if they could buy directly) - Obviously the benefit is that the RBOC only goes to one place for support - and you lose the "he said, she said" game when there is something wrong and vendors blame each other.

Next, some network vendors use the plugable optics as a way of making the "pay-as-you-grow" network work. You buy a 10 x 1G line card, and if you only want one port turned on, you only pay a relative fee. The way they prevent the customer from turning on all ports by buying cheaper / re-using other SFP's is to use the PROM to make sure only their SFP's can be used, or to implement some kind of license keying in the s/w. The obvious problem with the latter is control - Nothing physical provides any security, so ensure that the ports are being legally enabled, and have been paid for.

Thirdly, despite MSA's - no sane vendor would ever release a product to market without testing interop with a selection of SFP's and "qualifying" some. This does cost real dollars. In fact, one cannot be sure that Vendor X's LX SFP and Vendor Y's CWDM SFP interop in the same card as Vendor Z's SX module. Just because there is a standard doesn't mean it works - You need only to look at the standards debates that go on inside the IETF, IEEE and others to realise that the best standards are open to vendor interpretation.

It raises the point that if a customer wanted to stock a specific vendor's SFP's for sparing, they could ask all other vendors to adopt, and test, that same SFP to qualify it and allow it to be used. The downside for the customer is that someone must pay for the testing, and since the SFP vendor may be different for every customer, the network equipment manafacturers will either have to swallow the cost, or charge it onwards, nor will a vendor be able to sell "pay-as-you-grow" networks.

Just some thoughts for discussion.

turing 12/4/2012 | 11:12:05 PM
re: Use Our Optics, or Else! Sometimes electronics should be left to EE engineers, rather than academics who don't want to pay for what they get.

So I'm not an EE, but even my simple brain (and working at a vendor who's gone through the qualification of SFPs) knows the following facts in relation to this topic:
1) There is NO SFP standard, despite what Wodelet's NANOG presentation said. (he repeatedly stated it was an IEEE standard - it is NOT) There is, however, an SFP MSA, which is not at all completely tight, nor does anyone have to meet it completely. (there is no policing of this, other than the vendor)
2) There are more than 2 SFP manufacturers. I'm not talking OEMs. At my company we got at least 4 different original manufacturers for SX SFPs.
3) Of the 4 vendors we had in, each of them had different characteristics for heat, EMI, and optics (eye pattern, jitter, tx power, rx sensitivity). Even their plug-in physical design was different, in terms of number of holes in the metal, length of metal coverage, type of metal, as well as the housing it fit in.
4) Of the 4 vendors we had in, 2 of them we had to reject due to their heat and EMI characteristics. Not cost. Not volume. Just on characteristics.
5) Even though we buy from the 2 vendors we qualified, we screen the parts and yes, we do find some percent that exceed the characterisitics. That costs us money.
6) Some of the price of linecards may actually be embedded in the SFP price. In other words, Cisco may be able to offer cheaper linecards because they make some of it up in the SFPs. That pricing structure is actually good! Because that way if you only use 8 of 12 port card, you pay less than if you used 12 of 12, for example. If you insist on buying the SFPs separately, then Cisco will need to change the pricing structure. (I am making assumption that they price this way)

7) Had Lightreading asked a vendor (anonymously, not using their company name) before writing this article, they would have been told this info and could have provided a more objective viewpoint. (although to be fair to Lightreading, marketing guys for the vendors probably don't know this...)
opticalfuneral 12/4/2012 | 11:12:04 PM
re: Use Our Optics, or Else!
not to mention quality and reliability. Do you want a SFP that uses rejected laser pointers put together by children in China? Or one that was actually tested to see if it works before it leaves the sweatshop?
Ben Crosby 12/4/2012 | 11:12:03 PM
re: Use Our Optics, or Else! Depends how well it was implemented. I doubt it is a list of ID's - more likely either a range, or some kind of challenge/response. There appears to be circa. 2k of EPROM space. Vendors could use it how they see fit.

Physical_Layer 12/4/2012 | 11:12:03 PM
re: Use Our Optics, or Else! I was thinking exactly the same thing ... there is gonna be a business opportunity for someone to help the telcos fake out the IOS into thinking the part is a "legit" part.

Just curious - can anyone explain the details in this? Does the IOS have a database of serial numbers for SFPs (or other devices) that were shipped *to Cisco*? If so, I imagine it will be very simple for people to overcome this.

data_guy 12/4/2012 | 11:12:03 PM
re: Use Our Optics, or Else! If the customers had been concerned about the quality of other-vendors SFPs, then the customers simply would have purchased CISCO SFPs. No way they would have asked Cisco to lock-out their purchasing options.
This isn't anti-competitive, really, since Cisco doesn't make the things. This is just another example of a hidden tax/hidden fee. Hmmm, how can we improve profits without spending any development $$$s?
Long term strategy always is to lock the customer into your product, and what better way than to make sure that no portion of his investment is re-usable?
Get used to this kind of BS, once a company finds how to get $$$ for no investment, they'll never give it up.
What's next- a cisco switch that will only accept packets from a cisco router? Fords that will only run on Ford-brand gasoline?
I think I'll start a company that will re-program an SFP to any serial#-manufacturing code you want. I think the fellow that had to buy $300,000 worth of replacement SFPs would pay $50,000 for the service! (of course, someone would say this violates the DMCA, but I'd have quite a war-chest pretty quick!)
Most revealing to me is the fact that it would take 6 months to rev IOS and change this feature. Shows how bad Cisco's huge, monolithic approach to the op system was.
bajanetworker 12/4/2012 | 11:12:02 PM
re: Use Our Optics, or Else! Despite all the lamenting about quality and service, this boils down to one thing - MONEY!

Given Cisco's enormous margins, what does a customer get for buying a Cisco SFP? A price between 50% and 100% higher than they could buy the equivalent part for themselves. Cisco, Nortel, Extreme and Foundry add no value to the purchase of an SFP, GBIC or Xenpak. They don't design, build or test them. What they do, is qualify them in engineering, but indeed, I seem to remember spending more time testing competitors gear than modules when I was there.

Cisco has always looked to reign in customer choice as a business strategy which is why the defaults for many configurations in Cisco products are for propietary functions when equally useful standards exist. Make no mistake, many people at Cisco would love to start serializing packets to prepare for a day when . . .
materialgirl 12/4/2012 | 11:12:02 PM
re: Use Our Optics, or Else! This practice is as old as IBM. People made millions in the "PCM" (Plug Compatible Market). If you get it right, and can fight the FUD, you are in like flint.
data_guy 12/4/2012 | 11:12:01 PM
re: Use Our Optics, or Else! Its even simpler than tracking serial numbers.
When you set up to buy the plug-modules from finisar, etc. you tell them what manufacturer code to fill in. We have a prod. here that uses SFPs, and we just want to be able to tell modules we sold from modules we didn't, for warranty and RMA purposes.
We evaluated the idea of the product rejecting all optics modules that were not bought through us, but considered it a rotten thing to do to our customers.
Not sure if its a write-once-only PROM on the modules or if its flash.
Page 1 / 4   >   >>
Sign In