We spent some time this week listening to the unique perspective of a new broadband leader, in Sckipio, where marketing Supremo Michael Weissman, pointed us to a Broadcom lock-in, talked about over-performance in its chip trials, and its innovative method for connecting its G.fast lines to optical backhaul.
He paints a convincing picture of a two horse race with Broadcom, confesses that chip development between the two companies are neck and neck, but coming from a vastly differing points of view, and claims 5 hardware partnerships and 44 trial projects underway.
While the cloud represents the answer to most technology questions right now, many old telco-based operators, who to this day rely on copper connectivity in the last hop to the home, are virtually stranded, unable to bring scale cloud apps to those homes, without a substantial increase in broadband capability. And that’s why G.fast was invented in the first place.
We have heard the question asked many times, “When will telcos be able to close the gap with cable operators, in terms of how much bandwidth they can bring into any given home,” and the answer has mostly been a yawning silence. If you had not kept your nose to the grindstone and your ear to the ground in broadband, you could be excused for thinking that most telcos are ready to turn up their broadband toes, and die. And without broadband (okay and cellular), what have they got?
Our favorite example is AT&T which has virtually stood still in US broadband with around 16.4 million customer homes, since 2007, while Verizon has done even less well, and cable has eaten all the new business in the entire US market.
And if telcos have nothing new to offer, they won’t place many contracts, and perhaps that’s the reason why VDSL chip specialist Ikanos came out with figures last quarter which effectively say that it has one quarter of cash left, having around $12 million having burned about that much during the previous one. A curious alliance with Alcatel leaves Ikanos living on borrowed time and a further $10 million of borrowed money after the current quarter, hoping to turn its fortunes around, but without a G.fast option in sight.
Instead Ikanos has opened up the full 30 MHz profile for VDSL2 and is using it to claim that its Fusiv chips are the only ones which have actually managed to support 1 Gbps broadband speeds, over 100 meters of copper.
We have highlighted in the past the fact that Broadcom is in a fight to hold onto control of telco broadband – it has singularly failed, along with one of its major partners Alcatel, to offer the humble telco sufficient broadband relief, in sufficient timescales.
What is the point in bringing Vectored VDSL2, which promises nothing more than 100 Mbps, maybe 200 Mbps at a pinch, to a marketplace where the main telco rival in cable, has already deployed DOCSIS 3.0 which can be pumped up to 300 Mbps, and already features that as a headline speed in its marketing. Cable is also about to take the short step to DOCSIS 3.1, which shifts the goalposts into the cable Gigabit era?
The point is that every country represents a different problem. Some have fiber far closer to the home, some, have a variety of loop lengths, some have just invested in new cross connect cabinet infrastructure, others have all that cost ahead of them. So each Telco needs a different solution. Ikanos reckons that by mopping up those telcos who want to change relatively little, and stick with VDSL2 for a while longer, it can survive, but the story Weissman tells us that G.fast is turning out to be applicable in more of those instances than even Sckipio had previously imagined and where it works, he claims his RFPs are coming out 20% cheaper than Vectored VDSL already.
The first thing we asked Weissman was when was someone going to buy Sckipio, after all it has a strong relationship with Lantiq, and Intel has already acquired it.
“We are not the type to go it alone, look at the partners we have built up. Inside 5 years Sckipio will either be acquired or go public and then be acquired. We acknowledge that, but our job is to maximize the money that is paid for it. When I was at Coppergate the first offer to come in was for $60 million, but we held out and eventually sold it for $190 million. Sckipio has already had its first offers,” he said but would not elaborate.
“We have so far raised $27 million, and it only took up $5 million to get our first chip out of the door. Others have spent $100 million achieving that. The critical thing for G.fast is to reduce equipment supplier lock-in for the telcps, and also offer it in dual mode for a gradual upgrade with Vectored VDSL and beat the capex for fiber installation by at least $1,000 per household. And that means offering it without any customer truck roll, telcos have to be able to send out a CPE for a self-install.”
“Capex has been the main pain for telcos. The big question has been about lock-in and the optical solutions of Alcatel, Huawei and Adtran which do not talk to one another, so the supplier who wins at the head end can charge a premium, and can win all of the business downstream. If I can control the network processor, I can control access and the home network.
As we all got together to define G.fast, the telcos realized that they had to take control of the specification so that this was no longer the case, and companies like Sckipio had a vested interest to help them.
Weissman gives the example of how the G.fast plugs into GPON, or GigE or other backhaul, “The big issue is connecting to optical backhaul and the OMCI (OMT Management and Control Interface), the management controller. GPON from different vendors is simply not compatible.” So what the telcos have done is to try to eliminate the lock in by specifying Netcomf Yang and Openflow as the protocols for controlling the optical connection. And at Scipio our distribution point unit has no built in optical technology, so we connect with an SFP stick which handles the optical transport for me.
Netconf is a protocol defined by the IETF to “install, manipulate, and delete the configuration of network devices using remote procedure calls in XML format and Yang is its data modelling language.
Sckipio says it can adapt to the different Optical protocols using an optical SFP module, and we can make a single part that will scale, because we are not making a different one for each of the optical connections. I can work with 10 different telcos with each one having a different optical technology and still use the same DP unit.
This makes the telco problem sound a little like the set top lock-in that Motorola inherited from General Instrument. It was able to artificially raised the price of set tops for over a decade, while US cable companies designed workarounds and called on the FCC to set up Cable Card, separating the conditional access from the set top. Cable has gone a step further down the same track with many CableLabs standards and also with RDK.
So Sckipio sees itself as the White Knight in this equation, earning loyalty points by playing nice and being the alternative to Alcatel and Broadcom, and attracting many low end Taiwanese manufacturers to embrace its chipsets.
“The cost of our DP is coming out at 50% of the price of the competition, due entirely to this feature,” grins Weissman.
One design win that we have mentioned in the past was confirmed by Weissman, the relationship with Adtran, which in turn is the incumbent broadband supplier to Deutsche Telekom. So far Deutsche has not even launched exchange devices to trigger the millions of vector capable VDSL devices it has installed, and has made few public statements about G.fast, but the relationship with Adtran has continued after Adtran bought the broadband arm of Nokia Seimens, and Adtran has now formally embraced the Sckipio chips.
Weissman referes to the Broadcom strategy as “integrate and dominate” and we have seen this in the way Broadcom has added WiFi or Bluetooth or MoCA into existing chip designs in an SoC, and then sold it at a price so low that discreet chip designers cannot compete.
Weissman said, “The tsunami is coming for Broadcom. Proprietary designs will be pushed out of telco broadband architectures, and if Broadcom wants to get a piece of it, if it sees this coming, then it has to follow suit. Are they willing to deal? It’s not clear yet.”
Weissman tells us that AT&T has specified a virtualized optical environment, using NFV they have taken SDN from end to end, and specified their own management software for all their OEMs. If you have a virtualized environment then you don’t have to wait 2 or 3 years to get connection permission for a new part.
So will telcos have to go through Vectored VDSL to get to G.fast? “In the real world, in RFPs I have seen, the cost per device is already coming out 20% less than vectored VDSL.” Well he would say that. Weissman has no vested interest to push VDSL, and he forgets to mention that in most cases the DP has to be a lot closer to the home in G.fast, than in Vectored VDSL, so there can be considerable capex to spend over and above the per device cost.
But he makes a further point, “An operator doesn’t need 2,000 homes in an area to break even if the part is dual mode and can use VDSL chips too, he might do it on two customers.
Sckipio and others have rightly pointed out the MDU marketplace was the easiest to hit with G.fast in the first instance, but Weissman reckons that he has started going after many more customer types. “We are achieving twice the distance from the home that we expected. We have tested at 150 Mbps at 500 meters from the home, instead of the anticipated 250 meters.”
At first we told our sales team not to even bother with incumbent rural telcos, but it turns out that the performance of the chip means people are happy to install it further away. So far our reference design supports 16 devices connected to a single DP. Adtran’s supports 24 homes. Now customers are talking about 48 and maybe even 96 homes being connected and that may come out as a specification at some point.
“Bell South told us that at 500 meters they were getting maybe 30 Mbps, if you can get me 75 Mbps we are golden, they said.” He is suggesting he can get them twice as fast as that.
We question this, used to Weissman and others bolstering their claims unduly. “The main change is how quiet our chips runs. It has a very a low signal to noise ratio, far less lower electrical noise than we expected.”
The plan is still to go after MDUs first, and then take G.fast technology into the suburbs.