ATCA/Standard Servers

ATCA Finds Its Stride

WALTHAM, Mass. -- ATCA, AMC & MicroTCA -- Here's a sign of how AdvancedTCA (ATCA) has grown in acceptance: Some vendors want an ATCA chassis but intend to design their own blades to go into it.

It's starting to happen, according to Radisys Corp. (Nasdaq: RSYS) and others here at Light Reading's "ATCA, AMC & MicroTCA" conference on Thursday. Most speakers during the day said ATCA is really hitting its stride now that it's gone through a full generation of equipment.

Admittedly, it's taken six years since the PCI Industrial Computer Manufacturers Group (PICMG) ratified the first base specification for ATCA, which provides the building blocks for telecom equipment. (See Whatever Happened to ATCA?)

"The people who took the plunge four years ago maybe had a hard time. People who took the plunge two years ago have been pleasantly surprised," conference moderator and Heavy Reading analyst Simon Stanley told Light Reading between panel sessions.

The idea behind ATCA is for equipment vendors such as Alcatel-Lucent (NYSE: ALU) to save time and development costs by buying pre-designed chassis, blades, and/or middleware from the outside.

The concept butts against the in-house development that still dominates most equipment vendors, Cisco Systems Inc. (Nasdaq: CSCO) and Juniper Networks Inc. (NYSE: JNPR) being prime examples. And ATCA continues to face competition from HP Inc. (NYSE: HPQ) and IBM Corp. (NYSE: IBM), with their BladeSystem c-Class and BladeCenter products. (See HP Ditches ATCA.)

Stanley is predicting a $7.3 billion ATCA market by 2012. (See ATCA Takeoff Is Imminent.) That's a nearly 100 percent per year growth rate, a figure the ATCA vendors aren't challenging.

"We're starting to see more widespread adoption," said Ian MacMillan, senior product marketing manager at Interphase Corp. , during a morning panel. "The Tier 1 [equipment vendors] have been a little bit slow. We're starting to see the Tier 2 and Tier 3s build platforms around this technology."

And then you've got the vendors taking advantage of ATCA by using the chassis -- saving development time -- but developing all the blades themselves, meaning they get to keep the system proprietary. It's becoming a popular method in China, particularly with Huawei Technologies Co. Ltd. , one source at the conference said.

"Within the past year I've seen three customers do that," said Eric Gregory, RadiSys's ATCA senior product manager, speaking on a roundtable about the impact of ATCA.

You could consider RadiSys to be losing potential sales that way, but Gregory says the practice works out fine for his company. RadiSys even gets consulting business helping those vendors make the cards.

ATCA gear is still relatively expensive, and that's made some equipment makers shy away in the past. But MacMillan and other panelists said that's become less of a factor; the theory is that off-the-shelf building blocks save the equipment vendor money in the long run. Moreover, the premium gets shaved down if the volumes are high enough.

"The cost of chassis that have gone into production, that have really high volume, isn't a huge premium over doing it yourself. I'll put it in the 10 percent range," said Tom Ertel, vice president of engineering at Genband Inc. (Actually, that's the NextPoint Networks part of Genband -- see Genband Scores NextPoint.)

However, "when new blades come out, I do find it's a little bit cost prohibitive."

Interoperability among ATCA vendors has gotten a lot better -- Ertel noted he's never had a problem in that area -- but it's still not perfect. That means many ATCA projects remain sole-sourced.

Alastair Hewitt, software architecture principal of startup 4DK Technologies Inc. , noted that his company is sticking with RadiSys as its ATCA supplier. Paul Steinberg, Motorola Inc. (NYSE: MOT) fellow and chief architect, said his company had nine ATCA products shipping and had used two suppliers, but with no crossover; each product is sourced from only one of the two.

At least interoperability is better than it used to be.

"I remember the days when we were going to these plugfests and people were trying to figure out whether these things were compatible mechanically. Blades wouldn't fit into chassis, and people would take mallets to get them in. I'm not kidding," said Venkataraman Prasannan, RadiSys's senior director for ATCA products.

— Craig Matsumoto, West Coast Editor, Light Reading

Pete Baldwin 12/5/2012 | 3:31:34 PM
re: ATCA Finds Its Stride It's time to stop discussing the benefits of ATCA; those arguments seem well accepted (although the price is still a stickler).

Yeah, Cisco and Juniper aren't using it -- but 1) that's their culture, and 2) They were already working on next-gen boxes as ATCA emerged; if they use the stuff, it would be for the NEXT-next-gen gear. Wait and see.

So, ATCA wins, or at least it doesn't lose. What's the next big challenge facing systems designers?
goundan 12/5/2012 | 3:31:32 PM
re: ATCA Finds Its Stride Craig,

crApTCA has its place under the sun as long as it is used for purely computing purposes. When it comes to bit shovelling or anything that requires higher touch high speed packet processing or god forbid TDM/SONET from the dinosaur age, crApTCA lacks hardware arbitration logic, physical traces required for switch fabrics, thermal dissipation and board space, etc to do anything worthwhile.

crApTCA is great for brain dead companies such as LU, NT, etc which have gutted their hardware teams and are run by cluless idiots in management that believe everything under the sun can be implemented in software. Another factor in such companies is anything their development touches runs into the hundreds of millions and years to complete. Can't blame the idiots in mgmt completely either.

The purpoted benefits of crApTCA claimed by Chuck Byers will never come to fuition in your lifetime or mine. A limited set of those tall claims is definitely possible as long the application matches the capabilities of ATCA.
Pete Baldwin 12/5/2012 | 3:31:30 PM
re: ATCA Finds Its Stride Not a fan, eh? :) Well, you may have a point. We'll see.

Didn't Chuck Byers go to Cisco?
StartupGrunt 12/5/2012 | 3:31:06 PM
re: ATCA Finds Its Stride Yes I would say not a fan -- but not a fan of what? ATCA or Chuck?

As an ATCA seller, we always find that our major competition is not other platforms but instead the internal development organizations that want to do it their own way and are often positively offended by the idea that anyone but them can make good engineering decisions for their project. Sure ATCA has its shortcomings but I don't recall anyone significant as billing it as the perfect platform for all telcom purposes. If that had been anywhere near an industry consensus there wouldn't have been such a stampede to develop micro-TCA.

We have experienced some re-tooling of expectations for ATCA over the last several years. 20/20 hindsight tells us that it was never realistic to expect the market growth that was originally projected, and at the same time the rationale for the standard is as strong as it ever was. The maturing of this market segment is good news.
Pete Baldwin 12/5/2012 | 3:31:02 PM
re: ATCA Finds Its Stride Good post, SG, and it's true that internal development remains ATCA's biggest competitor.

I'm curious what you (or anyone else) see(s) as the next major challenge for ATCA. 10-Gig could be one answer, but I'm wondering if there are larger-scale, systemic changes that could benefit the platform. MicroTCA was one example - what's next?
goundan 12/5/2012 | 3:30:52 PM
re: ATCA Finds Its Stride Grunt,

ATCA's premise was and is low cost, easy development and 3rd party availability for telecom applications. As long as the application resides in control plane intensive server land, the premise is true. Once the application moves to bit shoveling land of any substantial type, the premise breaks apart really badly.

Even for applications where we move around 500 chassis a year, ATCA is 2x the cost and the kludges to get the system to working are absolutely horrendous. What is point of saving HW development dollars, when we spend more that what we save in SW development to develop, integrate and test kludges? If the answer is custom cards in ATCA form factor, then what exactly does ATCA fetch for us? Sheet metal and fans? That's absolutely the lowest cost item in any HW development expenditure if you know what you are doing.

ATCA does need to retool itself for bit shoveling land before it can make any claim whatsoever about being appropriate in such telecom applications. Call it ATCA 2.0 or ATCA++, but increase the board space 2x, increase slot pitch, switch fabric traces, remove the Phy from the control plane traces, reduce weight, change the architecture of the switch fabric / control plane interaction, add HW arbitration logic, change control plane redundancy architecture, etc. The rest of the stuff like e-keying, HPI, S/W, etc can be reused in the new platform with very little changes.
Sign In