AdvancedTCA Makes Headway

Advanced Telecom Computing Architecture (AdvancedTCA) -- an open standard for building telecom gear -- is gaining in popularity, judging by a new product announced today, on the eve of the CeBIT tradeshow in Hanover, Germany.

Telco Systems (BATM), an Israeli manufacturer of IP gear, is launching what it claims is the first IP-based platform built according to the AdvancedTCA specifications. The product, T6Pro, is a Layer 3 switch aimed at carriers and enterprises (see Telco Systems Intros T6Pro).

AdvancedTCA, also known as PICMG 3.x, is a family of specifications that defines a way of building the next generation of "carrier grade" telecom equipment with switching capacities up to 2.5 Tbit/s in a single shelf. The specifications contain enough information to allow board, backplane, and chassis vendors to independently develop products that will be interoperable when integrated together.

It is being developed by the PCI Industrial Computer Manufacturers Group (PICMG), a consortium of over 400 companies. More than 100 vendors are actively participating in the development of PICMG 3.x , making it the largest standardization effort in the organization's history. The key driver behind it is Intel Corp. (Nasdaq: INTC).

Motorola Inc. (NYSE: MOT), NEC Corp. (Nasdaq: NIPNY; Tokyo: 6701), and Siemens AG (NYSE: SI; Frankfurt: SIE) are working on wireless infrastructure based on AdvancedTCA, while Force Computers and Intel are developing computing-based platforms.

The idea behind the standard is that blades from one vendor could be incorporated into a chassis made by another vendor. "Today, if you look, you have a separate infrastructure for ATM, IP, wireless, storage, PSTN switches, and so on," says Danny Berko, BATM's product manager. "One day all services could be integrated into a single platform. The concept is any protocol on any card on any slot on an AdvancedTCA platform."

Whether this level of interoperability is a practical reality remains to be seen. Establishing standards and implementing them is a good start, but service providers will probably take a lot of convincing before they'll start mixing different vendors' cards in the same chassis. All the same, the existence of the standards will probably help drive down equipment prices, because it will lead to greater commoditization of subystems as well as the components that go into them (see Switch-Fabric Chipsets).

BATM claims to have sold its product to three major customers already -- the German Navy, the Israeli Army, and an Italian bank -- and a fourth customer in the U.S. is in the evaluation stage.

However, several key names in the IP world are missing from the AdvancedTCA standardization effort, most notably Cisco Systems Inc. (Nasdaq: CSCO).

"You will not see Cisco there, because it has a proprietary interface," Berko contends. The standards might weaken Cisco's ability to lock in carriers.

"While some still feel that hardware standardization will eliminate the ability to differentiate products, we disagree," says Ernie Bergstrom, principle analyst at Crystal Cube Consulting. "This is both a shallow and short-term view, and simply not true. Most vendors will keep their proprietary fabrics and backplanes until the architecture no longer scales to the required performance. The advent of ATCA may even shift the value proposition to the software domain where it belongs."

Cisco will be forced to jump on the AdvancedTCA bandwagon sooner or later when demand increases -- which it will, if recent market forecasts are to be believed. Indeed, RHK Inc. forecasts that the market for AdvancedTCA products will be worth $3.7 billion by 2007 (see ATCA to Be Worth $3.7B in 2007). Crystal Cube's forecast is more ambitious, predicting a market of $20 billion by 2007.

Cisco, for its part, says it is evaluating membership in PICMG. "Cisco subscribes to open standards interfaces leading to multi-vendor interoperability," the company wrote in an email to Light Reading. It declined to elaborate.

— Pauline Rigby, Senior Editor, Light Reading

sevenbrooks 12/5/2012 | 2:12:31 AM
re: AdvancedTCA Makes Headway
Or it could be that we actually build something that works, costs less, and that makes our products a winning propostiion.

My money is on that.

Compact PCI lets see volume applications - looks for it - nope sorry. Lots of niche markets. Still looking.

Keep repeating that Intel will take over the world and build on ATCA. Its a great idea, for me.

See you at the bank.

Sisyphus 12/5/2012 | 2:12:32 AM
re: AdvancedTCA Makes Headway > .. Standardized architectures have no value in
> volume markets for teloecom

Kee repeating it like a mantra so you keep believing it. You have not been able to remotely articulate a single reason why that is the case. Also, you're utterly wrong, there are already very standardized volume markets for telecom. For many years people have fit into 19 inch racks, just as one example. CmpactPCI rules several applications. ATCA just takes exactly the same philosophy to the same level, and it will agin significant volume *in* volume markets. It's niche markets that call for specialized solutions. You have it backwards.

> I know people that make their own framers (for
> example we make one that is half the price of
> the equivalent merchant part).

3 possibilities:

(1) It's only half as good.
(2) You dismiss the $M in development cost as operational expense, and merely compare the final ASIC price, which is laughably bad math.
(3) A combination of (1) and (2)

My money's on (3).
sevenbrooks 12/5/2012 | 2:12:38 AM
re: AdvancedTCA Makes Headway
That is correct. Standardized architectures have no value in volume markets for teloecom. I stand by that (and have been consistent about that). That is different that standard silicon. There is a market for it. For example, there is a Market for DRAM. That is standard silicon.

I know people that make their own framers (for example we make one that is half the price of the equivalent merchant part).

In little niche markets, go do ATCA or cPCI or whatever.

Sisyphus 12/5/2012 | 2:12:39 AM
re: AdvancedTCA Makes Headway
I am glad that you have backtracked a long way from your original premise. There were no parallel discussions. You categorically claimed modularity and standardized merchant architectural blocks have no place in telecom. Maybe it was for te sake of polarizing the argument as well, contrarian positions are fun to hold on these boards.

Even your latest example stands on very weak legs. I distinctly recall people saying that about POS framers in 95. Now who designs their own framers? This is high tech, yesterday's hero becomes tomorrow buffoon if he holds on to the same notion for too long.
sevenbrooks 12/5/2012 | 2:12:54 AM
re: AdvancedTCA Makes Headway
Perhaps you and I are having 2 discussions here. We are talking about an architecture that provides it all for you just write your application code. Just like the PC industry. That is flawed in telecom.

The place that standard semis fall apart is switch fabrics. They tend to be 10x what you can make them for. Even with the software costs. Line interfaces are pretty inexpensive but even there when an interface stabilizes, its attractive to go make your own (and save the 50% margin that you pay the semi guys).

Sisyphus 12/5/2012 | 2:12:54 AM
re: AdvancedTCA Makes Headway > Telecom Equipment has been around longer than
> computing and has had at least 5 attempts to
> do this. And none of them worked.

If you're talking about commercial silicon, you're wrong. Look at the evoultion of class 5 switches - classical example for a transition from maginficently specialized hardware to utterly commercial semiconductors, and the only differentiator becoming software. X.25 and early FR switches were built with utterly commercial CPUs, Intel's 80386 acted as a "NPU" in many of those. The list is never ending. The ASIC datapath is a recent development, and represents the exception in telecom. Your view suffers from severe 1995-2003 myopia, as if telecom and data networking had never existed outside of that timeframe. And even within that timeframe several netwqork applications are thoroughly standardized. Ever seen equipment based on CompactPCI and such? If not, it just comes to show you're fixated on the leave in the tree.

> Consumer products as well. Hunh. A single, 1,
> market has truly evolved the way you state.

You do know better, thus I am not even going to correct this. I guess semiconductor companies are an aberration, they just subsist on the far and inbetween part. There's over $100B that prove you're wrong on a yearly basis.

This has become a classic example for dicussions between pragamtism and ASIC fundamentalism.
sevenbrooks 12/5/2012 | 2:12:57 AM
re: AdvancedTCA Makes Headway

Telecom Equipment has been around longer than computing and has had at least 5 attempts to do this. And none of them worked.

Again for large scale markets the Opex associated with Engineering is interesting but certainly not relavent percentage of life cost. You can certainly do it for less R&D money, but at the end of the day what does the success of Alcatel and Adtran show us. BOM costs matter.

Lets take a look at a non-telecom, non-ASIC version of this. US Robotics and dial modems. What was their advantage? They hand coded (in assembly in many places) the modem controller into the DSP that ran the modem data pump. Thus, their BOM cost was less and they could make modems for less money. They made more money doing that than all the folks taking Rockwell modem chipsets and dumping them on boards, except for Rockwell.

So, its NOT just ASICs but they are a great and easy example. Its not inevitable. In fact, it has never been inevitable in the consumer space. Take a look at game stations. Lots of custom silicon there. Consumer products as well. Hunh. A single, 1, market has truly evolved the way you state.


sevenbrooks 12/5/2012 | 2:12:57 AM
re: AdvancedTCA Makes Headway
It gets to be more expensive in 3 ways:

1 - Silicon costs: With ASICs you don't pay markups on the silicon to a merchant vendor. For core functionality, this can be a 50 - 90% savings.

2 - Markups on all those other items. So, you get a chassis from somebody who designs it and gets a CM to make it. Markup again.

3 - Unoptimized. Systems are targeted. Generic systems are not.

Now what are they good for, lowering the cost of development of said items. When does that matter? When you are not going to sell very many. So, for low volume (niche) applications off the shelf is a fine thing. Costs will be lower for those applications.

Compact PCI was going to take over the world as well right. And Telecom Bus. and and and

Hias 12/5/2012 | 2:12:58 AM
re: AdvancedTCA Makes Headway How can ATCA, e.g. regarding rack, internal interfaces, be more expensive than proprietary equipment? I thought that ATCA stuff should become cheaper through competition on these issues.
Sisyphus 12/5/2012 | 2:13:03 AM
re: AdvancedTCA Makes Headway > .. It will be interesting to see say in Core
> Routing when Huawei gets its first real
> foothold in the US and starts selling its
> product at say 40% margin what will happen to
> Cisco.

They'll sell at 0% margin for as long as it takes todiscourage someone moving in there. They have the warchest to do it. I am not saying Cisco is invulnerable, but to challenge it in its key business areas requires b*lls of such size those executing on such a strategy better have wheelbarrows to cart those around in. Cisco can fight, and when it fights it is successful. Which doesn't mean they'll keep a 70% dominance going in all key markets, but they know it, and that's why they are diversifying more into new areas of growth constantly, and have an annoying habit of being successful in 90% of the things they take on.

I think Huawei will gain success if they execute right, but I don't thinl it'll supplant Cisco in the infrastructure anytime soon. The little gorilla will annoy the hell out of Papa Silverback, but if it tries to take over too quickly it'll get stomped into the ground. Huawei will pick its niches at first.
Sisyphus 12/5/2012 | 2:13:03 AM
re: AdvancedTCA Makes Headway The price of silicon is only fixed if you brush ASIC development under the carpet as basic operational cost. I mean, come on, this is the typical way it's done in networking: margin is caculated from the BOM exclusively, the whole R&D effort is not considered, and it's never quite clear whether it pays off or not. When you suggest it, people look at you as if you were a lunatic. Your R&D is a mysterious "fixed cost" you don't factor into product line profitability. Networking is shaky in understanding the true business mechanics to sustained profitability, and the factor "luck" is required for ultimate profitabillty when you account for products like that. Either it is a huge hit, or you lose your shirt. It is quite astounding, really. Merchant silicon allows one to truly calculate profitability, which of course several ASIC development cszars have little interest in. Eventually, in several markets they will get their stubborn posteriors handed to them by new entrants that execute at lower risk and lower R&D with merchant modular components, and still manage to get in the key differentiator.

The second commandment of ASIC religion is that it is also a given that you have to blow said "fixed" R&D wad on developing ASICs, which is why the software teams that are kept on the side are hugely overloaded - there's not enough left in the pot to truly invest into software development, and consequently I would very conservatively estimate that at least 50% of the ASICs' features developed in networking are actualy never used in system. It's always a software limitation given the fact those engineering teams allocate their $ into the silicon, and chronically under-fund the software development.

The thrid commandment of ASIC religion is to chronically confuse "differentiators" with "ASICs". It's like saying the key differentiator of car is "pistons". Customers don't give a hoot whether you do things in silicon or software, they just want it to *work*. What they want to see from system vendors is the ability to roll out services flexibly. To see the "key" in the silicon is to be fascinated by the leaves in the tree, and not even remotely comprehend the actual forest.

It will still take a while to see the transition, but it is inevitable. Always has been, and networking isn't about to be the exception to the universal rule of things that happen when technologies go into commodotization mode. Ten years ago, building packet forwarding ASICs that did things wire-speed was an art. These days teams that can do that are a dime a dozen, and the result is a lowest common denominator platform that is fast, but doesn't do anything particularly useful for the customer.

Cisco and Juniper's ASIC religion is total fallacy: they are successful thanks to IOS and JunOS, and *despite* their ASIC religion. Someone's bound to notice eventually.

And yes, this is a bit over the top for the sake of a polarized argument. as I said, things are not as one-sided, and will not change overnight, and nowhere have I claimed ATCA is going to be the ultimate slayer of proprietary platforms. It's only a first humble step in the right direction right now, but perhaps will initiate a stampede. Let's enjoy the show as it unfolds, it will be interesting to see what happens as world collide. I am not sure the networking world realizes what they have coming their way once blade server architectures truly deliver on some of the visions they have right now...
sevenbrooks 12/5/2012 | 2:13:14 AM
re: AdvancedTCA Makes Headway
Agreed you need to invest in what sets you apart (I think I said by standard stuff for things outside your core business). I think there are markets where COGS leadership is not the primary issue, however these markets tend to be niches (as I think I said before).

Bulk markets will, in the end, rely on COGS leadership. NREs in mass markets are irrelavent ($1M in NRE on 1 Million pieces/annum = $1/part in the first annum). It will be interesting to see say in Core Routing when Huawei gets its first real foothold in the US and starts selling its product at say 40% margin what will happen to Cisco.

particle_man 12/5/2012 | 2:13:16 AM
re: AdvancedTCA Makes Headway The price of Silicon may be fixed, but the cost of developing ASICs is extremely variable to the point it can sink a company. There is risk developing ASICs and it is a business decision whether to go that way or not.

I don't think the question is whether to build a differentiated product, it is what is the source of the differentiation.

In many cased it is not the backplane, chassis, or processor boards. So why not but those from some standards based source? That means lower development costs and cheaper COGS. Then put your scare R&D resources into what makes your solution different - maybe control plane software, performance, or even ASICs - whatever you have chosen as your key advantage.
sevenbrooks 12/5/2012 | 2:13:26 AM
re: AdvancedTCA Makes Headway
Sorry but it happens to be true. The cost of silicon is fixed. You can either pay margin on top of the cost or not. If its not core to the business, people buy off the shelf ICs and go on their merry way. If its the core cost driver, better off to own it and save 50% or more. We have a case internally where the ASIC is 1/10th the cost of the off the shelf IC. They do the same job and have the same gate count.

That is directly out of the business 101 book. The lunacy is thinking that one can take undifferntiated product to market based on off the shelf designs (say any Network Processor) and build something that is advantaged (see every startup in the 1998 and 1999 classes). Thats why truly successful products are so rare.

Sisyphus 12/5/2012 | 2:13:30 AM
re: AdvancedTCA Makes Headway > Guys, the reason that people do proprietary
> architectures is that they can be sold at a
> lower price.

This is repeated like a mantra in networking, but it's lunacy in the business 101 book. Yeah, the computer world sticks to proprietary chips like glue. So do consumer products. NOT. They rely on off the shelf components, and a modular architecture with clear standards. It's supply chain basics. Networking still likes to think it represents some sort of niche market, like building exploration pods for NASA, and suffers from NIH syndrome that affects cost points and hurts it, badly.

Alas, the future is inevitable. Good ole 80/20 rule: 80 percent of networking systems will be based on utterly standardized and modular building blocks, and a true 20% niche will not. In fact, the adoption of such dramatic cost reduction measures might fuel a new upturn, as it will deliver benefits to everybody.

Which standard it will exactly be I don't know. I am not sure ATCA truly gets everything right to cover 80% of applications, but it's all a learning process and we'll get there within 2 years or so at the most.
sevenbrooks 12/5/2012 | 2:13:37 AM
re: AdvancedTCA Makes Headway

Think of it this way. Silicon costs in any moderate kind of volumes are somewhat fixed: i.e. a 1.5M gate, .18u chip cost is about the same whether you make it or somebody else makes it. If somebody else makes it they amortize their NRE and then add their margin to the cost. If you build it then you amortize the margin only.

Every time there is a layer (chips, modules, boards, systems) added to this one ends up in margin stacking. So, that means you are paying the gross margin for each layer.

Now lots of DSL makers can make DSLAMs at $50 per port and thats because there is a merchant market for the silicon (although its not directly replaceable) and people have spent money in creating low cost architectures.

Standards are fine things taken in moderation. Its impossible for 1 standard to do all things or to do anything than its direct intent efficiently. So then you end up with the other problem of a 10 Gb/s system for a 1 Mb/s problem. The cost mismatch kills you.

Lets go to the PC version of this. Now in the old days (by cracky) the software to do e-mail, word processing, and spread sheets could run on a 10Mbyte HDD and 256K of codespace. Suppose one built a system that was optimized to do that job today. Could we build it for $100? So, why do we pay $1500 - $2000 for Admin PCs? The software has gotten out of hand and requires the $1500 infrastructure to do the $100 job. So, to me the PC has always been the model of inefficiency.

opticalwatcher 12/5/2012 | 2:13:42 AM
re: AdvancedTCA Makes Headway "Does Cisco charge for SW upgrades to 6500 series routers, or do they always manage to pacage a new HW card along with it?"

Yes, it's called a yearly "maintenance" charge. If you don't pay, you don't get the new software.
glasstotheass 12/5/2012 | 2:13:44 AM
re: AdvancedTCA Makes Headway
I work for a major telecom systems vendor. My view is that the real problem of "shifting the value proposition to the software domain where it belongs" is that that's not how our customers think. They want to pay for hardware ("how much does it cost me for N ports?") and not for software upgrades.

Our SW/HW engineer ratio is nearing 5:1, yet our customers won't pay for code upgrades. Supposedly, this is the case with most other vendors as well.

Does anyone know if there's movement in the industry to correct this imbalance? Things like TCA won't take off until it does.

Does Cisco charge for SW upgrades to 6500 series routers, or do they always manage to pacage a new HW card along with it?


fiberous 12/5/2012 | 2:13:51 AM
re: AdvancedTCA Makes Headway Particle,

Sure it would be cheaper in COGS because you
wont have any. Since, you wont be able to
sell any. Because the all dont go into
the chassis and sing a Xmas carol - mostly
they lokk back at us like a bunch of frogs
ina pond.

Look, these days most system companies have a
hard time getting their HW guys to work with
their SW guys to come out with things in time.
Imagine 15 HW teams, each from different
companies and SW from another, NMS/OSS
from another - all this go into the funnel and
and becomes rocket fuel.

particle_man 12/5/2012 | 2:13:53 AM
re: AdvancedTCA Makes Headway Well, maybe a different way to look at it is one step up the food chain. Imagine that I was nuts enough to start a god-box company. From a purely hardware perspective, (I will conveniently ignore the software and protocol mapping nightmare to make my point :-) I could go to the emerging ATCA market and OEM the backplane, the 12 or more different interface blades, some processor cards, and a couple of switch fabric cards. This would give me a pile of development leverage and (if SW wasn't critical path) improve my time to market.

Would it be cheaper from a COGS point of view? It may be.
fiberous 12/5/2012 | 2:13:54 AM
re: AdvancedTCA Makes Headway "The idea behind the standard is that blades from one vendor could be incorporated into a chassis made by another vendor."

What a crock !!

Clueless people - wake up!!
rtfm 12/5/2012 | 2:13:57 AM
re: AdvancedTCA Makes Headway sevenbrooks:

I am trying to understand. Don't standards allow more components to compete, as well as higher volume? One only has to look at WiFi (witness Intel into telecom, overall) to see this.

If Alcatel can sell DSLAMs for $50/port, is that not because of volume and amortized development costs? How much is becuase of their architecture/technology?

I've just been a believer of standards for bringing down costs, so trying to learn. . .

particle_man 12/5/2012 | 2:13:57 AM
re: AdvancedTCA Makes Headway That's an interesting point about the $50 DSLAM. I would think that the ATCA standard would be more suited for higher end elements. It's got all that high speed signal capacity on the backplane which would make it expensive for low speed applications.

You can definitely cut costs to some extent by specializing your hardware. You can also cut costs through standardizing on a high volume platform. The extreme example of course is the PC. I haven't done the math to estimate where the crossover point is but I think the open question is when ATCA will hit critical mass and create a compelling cost advantage.
sevenbrooks 12/5/2012 | 2:14:04 AM
re: AdvancedTCA Makes Headway
Has apparently never sold anything to a large carrier.

I am surely going to bet my company's future on what they have to say.

Guys, the reason that people do proprietary architectures is that they can be sold at a lower price. So, depending on what you make you may or may not be able to use a ACTA (or cPCI or whatever) standard platform.

Lets use a little example. Lets see, Alcatel DSLAMs. Go make a DSLAM thats say $50/port out of an ACTA. Call me when you have done that.

Until then, these will be nice little niche fillers (just like all the versions before them - Telecom Bus as an example).

edgecore 12/5/2012 | 2:14:05 AM
re: AdvancedTCA Makes Headway A- If ATCA is showing a 3.7 B potential market, and Force is hot and heavy into this space...then why is Selectron so keen on selling Force?

B- Cisco has always developed custom software for custom hardware...not too sure why the need to talk about Cisco in the context of open standards!

Sign In