The first Advanced Telecom Computing Architecture (AdvancedTCA)* systems are now hitting the streets, and initial indications are that this technology could be a real winner with telecom OEMs and carriers alike. As with many new technologies there is plenty of hype, but in this instance there is already real substance behind the headlines.

ATCA promises to bring "plug-and-play" to carrier-grade systems by defining a high-availability, chassis-based platform that will scale from gigabits- to terabits-per-second. ATCA systems can be used across many applications, from technology demonstrators to business-critical solutions.

This report takes a look beyond the headlines at the likely benefits, the applications, and the outline specifications, while exploring some of the recent developments supporting this new platform.

Here’s a hyperlinked summary of the report:

Background Reading

A Webinar on this topic, entitled "ATCA: The Interoperability Revolution Continues," is scheduled for December 8. It will be moderated by the author of this report, Simon Stanley, and sponsored by Intel Corp., Motorola Inc., RadiSys Corp., and Siemens Communications Group. For more information or to register, please click here.

Click here to view the recent Light Reading Webinar on which this report is based. The Webinar was sponsored by Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC), Intel Corp. (Nasdaq: INTC), Motorola Inc. (NYSE: MOT), and Telco Systems (BATM).

Need to know more about the latest developments in AdvancedTCA? check out the coming Light Reading Live! conference:

AdvancedTCA – The Architecture of Tomorrow's Telecom Systems
at the Hyatt Regency in Burlingame, Calif., on Thursday, January 27, 2005

This one-day event, hosted by Simon Stanley, will provide qualified attendees from Light Reading's global audience with original research into the ATCA components market, from its Heavy Reading market research division.

* AdvancedTCA and ATCA are registered trademarks of PCI Industrial Computer Manufacturers Group (PICMG).

— Simon Stanley is founder and principal consultant of Earlswood Marketing Ltd. He is the author of several Light Reading reports on communications chips, including Packet Switch Chips, Traffic Manager Chips, 10-Gig Ethernet Transponders, Network Processors, and Next-Gen Sonet Silicon, and is a regular moderator of Light Reading Webinars. He is also author of a Heavy Reading report entitled 10-Gbit/s Ethernet Components: A Heavy Reading Competitive Analysis.

During the late 1990s and early 2000s the focus of network OEMs was the delivery of high-performance technology-leading systems to meet the demands for ever greater telecom bandwidths. Chassis and backplane design were key differentiators for market-leading manufacturers. Although several companies investigated the market for off-the-shelf telecom blades, the cost of optimizing these designs for each OEM was a significant barrier to success.

In contrast, the market for personal computers and servers has moved completely to standardized designs. The use of standard chassis is also commonplace in industrial systems where CompactPCI, developed by the PCI Industrial Computer Manufacturers Group (PICMG), has largely replaced the VME bus. CompactPCI enables manufacturers to use PCI-based components from the computer industry in more environmentally demanding applications. However, CompactPCI cannot meet the needs of the telecom industry for high-availability platforms with a wide range of system capacities from Gbit/s to Tbit/s.

The recent downturn in the telecom industry has had a major impact on the size and shape of the companies supplying this market. Several have undergone major refinancing and restructuring, and most have significantly reduced their in-house development capability. But, as capital expenditure by carriers starts to grow again, the focus has moved to return-on-investment rather than absolute performance, so networking OEMs are now moving quickly to introduce solutions to meet this requirement. The need for a standard chassis solution supporting high-availability applications has never been greater.

ATCA-based systems can deliver significant benefits to OEMs and carriers when compared to designs using proprietary chassis:

Faster time to market: ATCA enables significantly faster TTM through the use of off-the-shelf system components and standard semiconductors. Developers can focus their engineering resources on the one or two blades that contain their own differentiating technology, and so use third-party components for the chassis and complementary blades required to complete the system. When carriers have a unique requirement, they have the option to source the core system from one vendor and add compatible blades from another.

Greater vendor choice: ATCA lowers the barrier to entry and will significantly increase the vendor choice available to both carriers and OEMs. With a modular platform and wide industry support, carriers are not necessarily committing to a single vendor when they buy into a solution. This gives carriers the opportunity to make purchase decisions based on current requirements, with less regard to future needs, opening up new, short-term opportunities for suppliers.

Increased flexibility: ATCA has been designed to deliver significant flexibility through a choice of processors and interfaces, both internal and external. The same chassis and management blades can be used across a wide range of systems, thus reducing inventory, while each system can be optimized for the specific application with different line and switch blades.

Reduced cost of ownership: Reduced COO is key in the current telecom environment, where maximizing return on investment is driving all business decisions. The use of ATCA significantly reduces the cost of R&D required to introduce a totally new system. The lower investment and the manufacturing savings from using a single chassis across multiple markets and customers will together lower support costs while driving down the total cost of ownership.

ATCA has been developed by PICMG in conjunction with the telecom industry, and is a significant development over CompactPCI and other Eurocard form-factor solutions. ATCA supports larger boards, higher bandwidth, and the higher power required for telecom systems. ATCA also mandates advanced system management and high availability. Legacy CompactPCI and PCI boards can be supported by ATCA through the use of carrier boards that adapt them into the ATCA environment.

The market for standard chassis-based systems, and ATCA in particular, is set for significant growth, driven by the need of both carriers and OEMs to take advantage of the benefits laid out in the previous section.

“We think the communications industry today is really embarking on a path towards modularity very similar to what you saw the server world go through in the early 90s,” notes Rajeev Kumar of Intel Corp. (Nasdaq: INTC) and president of the Advanced Switching Interconnect SIG (ASI SIG). ”We think some of the key enablers for that modularity in the industry will be AdvancedTCA and Advanced Switching technology.”

Figure 1 shows the likely impact of this change. In-Stat/MDR has forecast that 15 percent of capital expenditure (capex) on chassis will be spent on systems using a standardized chassis by 2008. This is up 15-fold from the 1 percent spent on standardized chassis in 2003.

For ATCA in particular, RHK Inc. has forecast the market to reach $3.7 billion by 2007, while Crystal Cube Consulting has forecast this market to reach $20 billion in the same time frame. Whichever figure is nearer the mark, the amount is substantial.

In a recent poll, taken during a Light Reading Webinar on ATCA, 70 percent of the audience were either using ATCA or planned to do so by 2007.

“We see nine out of 10 Tier 1 [telecom] OEMs developing ATCA-based systems,” says Gilles Garcia, director of marketing for switch fabrics at AMC Corp.

Initial Applications

ATCA has the flexibility to address high-availability applications across many business-critical functions. However, in the short and medium terms, the deployment of ATCA systems will be mainly confined to particular market segments.

In the short term (2004–6), the deployment of ATCA systems is likely to be in medium-volume, packet-based applications with throughput limited to a couple of Gbit/s per slot. These initial applications will include:

Other ApplicationsATCA opens the door to hybrid systems combining several of these functions into a single, customer-driven solution. Network security will also be a growing application for ATCA, with encryption and intrusion detection being incorporated into many carrier systems. In the medium term, ATCA should see significant deployment in the carrier access and edge, including B-RAS and DSLAM, as the broadband network becomes more IP-centric.

In the longer term, ATCA may also be used in the carrier core and transport as well as within the enterprise. These high-bandwidth applications rely on higher-performance fabric interfaces that are still in development. These interfaces are the source of many debates, with several companies developing proprietary solutions ahead of the full standardized specifications.

“We see a growing demand for ATCA-based solutions for carrier access applications,” says Avi Cohen, VP for research and development at Telco Systems (BATM). “We believe Far East deployments will pave the way and push forward the carrier applications into the short-term period.”

ATCA is a common platform for high-availability telecom and computing applications, and is ideal for central-office applications that require ETSI and NEBS compliance.

The PICMG3.0 specification defines the ATCA base requirements. For higher-bandwidth applications there are the PICMG3.x subspecifications that define application-specific fabric interfaces. The first of these is PICMG3.1, which defines the Ethernet and Fibre Channel fabric interface.

Key objectives in developing ATCA as a common platform were that it should have:

Shelf Management

ATCA systems must include shelf management. This monitors and controls the ATCA boards and other field-replaceable units. Shelf management also controls the power, cooling, and interconnect across the system. This function is particularly important during failover to back up components and during the in-service installation of new components.

“Shelf management and high-availability management are key to ATCA,” notes Mathias Renner, strategic marketing manager with Motorola Inc. (NYSE: MOT). “It is more complex than the previous generation, and so it is taking time to roll out products.”

An ATCA system may include dual redundant shelf managers, and, like the rest of ATCA, this function is interoperable across multiple vendors.

ATCA Chassis

The ATCA chassis is designed to fit the latest 600mm ETSI shelf; however, it can also be used with older 19" or 23" shelf systems.

Figure 2 shows the basic ATCA form factor, with the front of the chassis on the left and the rear of the chassis on the right. With the front or main board is an 8U-high 280mm-deep blade, connected to the backplane through interfaces in Zones 1 and 2.

Zone 1 supports power and control, and Zone 2 supports the base and fabric interfaces for data transport. Power to each slot is +/–48V with up to 200W per slot. The base interface supports triple-speed Ethernet, and the fabric interface is application specific. Both interfaces are optional.

The ATCA backplane supports two to 16 slots. Further flexibility is added using the rear transition module (RTM) and mezzanine cards. The RTM sits behind the front boards and can be used for additional functionality or connectivity. The front board is connected to the rear transition module via Zone 3, but the base specification details no interfaces or connectors for Zone 3. Up to four fixed mezzanine cards can be plugged into a carrier front board, allowing ATCA to support smaller cards such as PCI and CompactPCI or interface-specific modules. Future ATCA systems will support hot-swappable advanced mezzanine cards (AMCs).

Zone 2 is used for data transport and is split into two interfaces: the base interface and the fabric interface.

The base interface provides 10/100/1000Base-T Ethernet connectivity across all the slots. This interface has a dual-star topology and can support low bandwidth applications up to 1-Gbit/s per slot. For higher-bandwidth applications the fabric interface can be used as a secondary communications channel in parallel with the base interface. In most high-speed networking applications, the base interface will be used to carry communications between the control-plane processors on each linecard.

Fabric Interface

For applications requiring a non-Ethernet backplane connection or bandwidths above 1 Gbit/s, ATCA includes the fabric interface. This is protocol agnostic, providing eight bidirectional serial interfaces per slot. The ATCA subspecifications PICMG 3.1-3.5 define several networking technologies that can be supported by the fabric interface:

Switching Architecture

The fabric interface is a flexible switching solution supporting both mesh and star topologies (see Figure 3).On the left of Figure 3 is a dual-star topology with centralized switching and redundant switch fabrics. This is the same topology as the base interface and requires a simple interface on each linecard.

On the right is a full-mesh implementation. Here the switching is distributed across the linecards, and the switching function on each linecard must be able to communicate with every other slot in the system.

Figure 4 shows a typical, ATCA-based telecom system with dual-star switching and Ethernet fabric interface. Only the front boards and the backplane are shown. The system has two switch cards and 14 linecards. Running across the backplane are the base and the fabric interfaces as well as the sync clock and update channel.

ATCA provides a common platform for high-availability telecom and computing applications. Using ATCA, OEMs and carriers can benefit from the same chassis for a range of high-availability applications.Three typical implementations are:

In the short term, the majority of systems deployed will be in server and wireless networking applications.

The modular system is more than just a hardware platform, however, and needs an operating system and application software. Many of these functions are now available as open-source applications.

Figure 5 shows a typical hardware and software stack for a server application. At the bottom is the ATCA hardware platform, while above is the carrier-grade Linux operating system. At the top are the customer applications through which a systems OEM can provide significant product differentiation.

The ATCA base specification was approved in December 2002, and over 28 members exhibited ATCA products at Supercomm 2004. The PICMG has more than 600 members, and many now offer ATCA components.

“The ATCA specification is complete, and systems are ready for deployment,” says Motorola's Mathias Renner.

Several semiconductor companies, including Freescale, Intel, and NPU vendor EZchip Technologies, have already adopted ATCA as a platform for technology demonstrators and silicon evaluation. As other companies follow suit, ATCA systems will be appearing in development labs across the world.

Components and systems available now include: