Your browser is not supported. Please update it.

22 October 2007

Consolidation indicates that MicroTCA is going mainstream in wireless

The build-out of 3G and post-3G data-focused networks, with their requirements for large numbers of base stations, is proving one of the most powerful drivers for adoption of the MicroTCA standards, which define sets of reusable building blocks to reduce cost and time to market for many types of telecoms equipment, notably wireless base stations. A smaller offshoot of the ATCA (Advanced Telecom Computing Architecture) standards, MicroTCA is starting to appear in systems for WiMAX and 3G, and is being combined with other emerging standards ‘ from base station interconnects like OBSAI to software platforms like IMS ‘ to offer vendors and operators a new set of economics as base stations start to commoditize. Amid these developments, some of the large vendors that led the charge to ATCA and MicroTCA are stepping back and selling off their actual products to more specialized suppliers, as the sector looks to gain volume but also come under heavy price pressure. Recently, consolidation was significantly accelerated by Intel’s sale of its ATCA activities to Radisys, and Motorola’s to Emerson.

One of the frontrunners in MicroTCA for WiMAX has been Emerson Network Power, through the Embedded Computing unit it built around its acquisition of Artesyn Communication Products, and now Motorola’s embedded activities. At the 3GSM show in February it previewed a MicroTCA WiMAX base station, with a target price of just $500 for a baseline box once the market hits volume. Emerson’s embedded system was demonstrated with an Intel WiMAX baseband processing card, E1/T1 and Gigabit Ethernet backhaul cards, and a system controller card.

Last month, Emerson Electric bought Motorola’s Communications Computing group, a venerable player in the embedded market, for $350m in cash. Combined with Artesyn, this positions Emerson strongly for a market estimated to be worth $6bn and growing, but where scale will beall-important as prices come under intense pressure. The Motorola unit had sales of about $520m in its last financial year while Artesyn revenues were about $100m. These will provide embedded computing products and services to communications equipment manufacturers in telecoms, medical imaging, defense, aerospace and industrial automation, complementing Emerson’s core activities in
embedded computing and embedded power. Motorola entered the embedded communications market in 1982 when it bought Four-Phase, but is now looking to retrench to its core businesses. Like Motorola, Emerson has been a major proponent of ATCA and MicroTCA. Emerson will not necessarily gain significant business with Motorola itself ‘ the company always sold its embedded products only to OEMs, not carriers, and claimed that its own division had to compete for sales to other Motorola divisions on a level playing field with rivals like Radisys. However, it may now be easier to sell to other major OEMs without the fear of conflict of interest with Motorola.

The Emerson deal came only months after Intel decided to sell its PCI and ATCA board business to Radisys for $25m plus $6.75m of inventory and other considerations. Radisys expects its acquisition to add $50m per year to ongoing revenue, plus retaining a close alliance with the chip giant. “Intel is sharpening its focus on communications and embedded market segments that are aligned with our core businesses,” said Doug Davis, general manager of Intel’s Embedded and Communications Group. “The modular communications platform business presents an attractive market opportunity, and these products will remain an important part of the Intel ecosystem. We will collaborate closely with Radisys to ensure that customers have complete and competitive Intel Architecture-based solution.’

The brace of transactions means that the most prominent supporters of ATCA have now abandoned their own products, though Intel is adamant that this does not show a loss of confidence in the sector, but follows a common pattern for the chip giant ‘ to lend its weight to establishing a standard that will bolster its core products and its overall view of the direction technology should take, but then to hand off its own efforts to a specialist as the market takes off (it took a similar line in Risc processors, Multibus and network processing). Intel will, of course, retain an interest in ATCA as it will hope to sell its microprocessors to boardmakers, without the conflict that arose from having its own boards on the market (boards that never achieved high sales). Many customers who did buy Intel ATCA boards will welcome the transfer of the product to the more effective Radisys. An example of this is Intel’s Glenfield WiMAX ATCA board, which is generally thought to be well designed but poorly marketed, and so could be a more powerful product in Radisys’ hands, targeted at vendors seeking to enter the WiMAX market quickly and with low risk. In the early days of WiMAX, Intel had looked to develop a full base station platform itself, but then shifted its position and created Glenfield, which combines its own MAC with the physical layer from picoChip. Radisys is expected to accelerate the upgrade of the product to the next generation of picoChip processors.

MicroTCA is poised to become more important in the wireless market, at least, than its bigger brother, since it is geared to high volume, low cost equipment. Both standards are controlled by the PCI Industrial Computer Manufacturers Group (PICMG), which sees wireless as one of the most promising growth markets for its members. They define generic telecom chassis, so that vendors can buy a shell of hardware and some software off the shelf and look for competitive edge in software. MicroTCA makes heavy use of advanced mezzanine cards (AMCs), standard modules with flexible interconnects that make them hot-swappable. Simon Stanley, author of a report on the platform for Light Reading, sees WiMAX as a sector where there will be early adoption of MicroTCA ‘ as there has been of OBSAI ‘ because vendors are designing products from scratch and can take advantage of new developments immediately. “WiMAX systems require high availability with hot-swappable modules, extended temperature range, and a cost effective solution. MicroTCA can meet all these requirements,” Stanley said. Other likely targets will be IP PBXs, 3G base stations, and DSLAMs, and even non-telecoms markets like image processing.

With MicroTCA standards finalized, most tier one vendors now have clear roadmaps and are being followed by tier two suppliers, and there are now about 30 companies offering MicroTCA and/or AMC, with commercial systems available from Emerson and Alliance, among others. Most MicroTCA systems support between two and 12 AMCs, which support various functions including control plane processing, digital signal processing network/security processing, storage I/O and specific applications. Vendors include Radisys and Interphase. The same AMCs designed as blades for full-blown ATCA systems can be reused as components of MicroTCA products.

ATCA, and now MicroTCA, are scoring well in wireless because of the trend to advanced networks with large numbers of smaller base stations, and because wireless core networks have less onerous speed demands than wireline systems ‘ ATCA is only now reaching 150Gbps capacity, and will not achieve 500Gbps until it has a 10Gbps backplane, in2008 or later. At this point, it will become a key platform for major converged core technologies like IMS.

MicroTCA allows reuse of common components such as chassis, shelf control, power, and CPU or I/O cards, creating economies of scale. More than half a WiMAX or 3G base station’s requirements can now make use of common components, with baseband processing and radio remaining the chief exceptions. A MicroTCA-based product in this category can cost as little as $8,000. MicroTCA also increases the efficiency of some base station attributes such as clocking and synchronization, where a reference clock must be distributed to all the baseband processors and radios. MicroTCA supports three general purpose reference clocks in its backplane design, which can be redistributed to all card positions.

The combination of ATCA and MicroTCA with other standards like OBSAI starts to fulfil the promise of a truly open, commodity base station with PC-style economics ‘ a trend equipment makers are increasingly willing to embrace as they seek to reduce R&D costs and as they gain less differentiation through hardware design, and more from software. OBSAI has devised a set of open internal interface specifications for base station architecture and module specifications. Its work covers the areas of transport, control, baseband and radio (see Wireless Watch January 30 2006). Early adopters include Nokia Siemens, which supports OBSAI in its Flexi Base Station range for WiMAX and 3G. The OBSAI architecture maps directly to the MicroTCA architecture and some vendors have even proposed that OBSAI specify MicroTCA as its preferred platform for open base stations.

Many software and firmware would-be standards are now pushing to be combined with ATCA and MicroTCA physical layers to become key parts of the evolving open telecom platform. The groups supporting these various platforms are loosely uniting, via the Mountain View Alliance, to combine efforts for ATCA and make it more attractive to OEMs. Among them are the Service Availability Forum’s Application Interface Spec for high availability, as heavily backed by Motorola, Ericsson, Nokia Siemens and Hewlett-Packard; the Communications Platform Trade Association (CP-TA), which focuses on interoperability; Scope Alliance, which analyzes existing profiles and specifications for specific telecom applications.; and OpenSAF, providing a working and open source implementation of SAF standards. Motorola has led the charge to make as many of these software layers as possible open source to accelerate uptake of ATCA and MicroTCA.

Another important factor in the creation of low cost, commodity base stations will be reduction of the cost of power. Most base stations currently have an operational power draw of 3 kilowatt-hours but a great deal of energy is wasted on the cable link connecting the radio gear to the antennas, leading to the need for large power amplifiers. This issue is being addressed by some vendors with remote radio heads, where the radio-to-antenna link is reduced from 50-100 meters to one meter or less. This is achieved by placing the radio transmission gear on the tower with the antennas, with input data to the radio delivered digitally over fiber. This can reduce the use of electrical power to 10% of current levels.

Siemens was one of the first proponents of remote radio heads, which are now also supported in the Nokia Siemens flagship Flexi and by many specialized vendors such as Celestica. Other advantages of this approach are that, with radio components at separate locations, 3G networks can be built with less redesign of existing GSM systems; and while the radio head and antenna are installed as usual on a roof or a tower, the server can be located in a basement, for example, where the site and installation costs are far lower. NSN says this can reduce site rental cost by as much as 70%, physical deployment by 65% and power consumption by 50%. One server can serve up to 12 radio heads, reducing the cost of network deployment.

Of course, the logical extension of the shift to open base stations built PC-style around off-the-shelf components is the full development of the much vaunted software defined base station. In this model, custom hardware becomes almost non-existent, and a standard, low cost hardware platform can support any radio technology interchangeably. All the baseband processing is then carried out on generic computation components such as NPUs, FPGAs and DSPs.

The boom in communications servers:

The move by telcos to create converged wireline/wireless networks delivering a wide range of multimedia services will put huge strain on their existing computing infrastructure, and the server makers are eyeing with rising excitement the coming round of hardware upgrades by the major carriers. Just as the enterprise vendors are joining the traditional telco suppliers on the software side of convergence, bringing skills learned in the corporate environment to the carrier market and platforms like IMS (IP Multimedia Subsystem), so they are jostling for position in the market for a new class of telco-optimized servers, and the battle to control the standards is already underway.

Unoriginally labelled Communications Servers, these are being rolled out by IBM, Sun and Hewlett Packard from the traditional server world, and by telecoms equipment makers like Motorola. They all have in common certain key technologies associated with multimedia service delivery and high availability, and with the suppliers feeling the heat of the intense price sensitivity that now affects the carrier market, they are increasingly based on standardized components, a new development in this sector ‘ and one that is throwing up a battle of influence between Motorola and Intel.

A Communications Server is an embedded computer that contains the following combination of features:

‘ Open standards-based common platform
‘ High availability (more than 99.999%) and serviceability
‘ Support for value add communications applications
‘ Advanced multimedia, routing and packet processing capabilities
‘ Extended lifecycle

Wendy Vittori, general manager of Motorola’s newly acquired embedded communications group (now part of Emerson), recently summed up the need for a new class of carrier-grade equipment. ‘To keep up with the stringent demands multimedia applications place on their support servers and networks, a new class of embedded computer is needed ‘ the communications server. Just as the enterprise server and enterprise computing emerged 20 years ago to address a major market need, we believe the communications server and communications computing will emerge and evolve for the same reason.’

Motorola and MicroTCA:

Emerson inherits a strong position in MicroTCA with its purchase of Motorola’s embedded communications unit, but also some political baggage in the standards community. While Intel, along with its close ally Alcatel, have been the strongest cheerleaders for ATCA, Motorola aimed to take the leading position with MicroTCA. As always, there was a strong whiff of vendor politics between Intel and Motorola, which have very different agendas for the next generation network market. While Intel is pushing PCIExpress as the switching fabric for ATCA, Motorola is supporting Ethernet and is aiming to steal a march on the chip giant in MicroTCA, an implementation for smaller telco systems, which should be Intel’s natural home.

So while promising full support for the key standards involved in a Communications Server ‘ ATCA, SA Forum and also Carrier Grade Linux, the most usual operating system for telco servers ‘ Motorola claims to have set two standards directions of its own with the MicroTCA form factor and the use of Ethernet in ATCA (other contenders to be the primary fabric include Infiniband, and arguments over this aspect of the technology have been a major stumbling block to uptake so far.) Motorola is convinced that affordable high availability telco servers will only emerge quickly if the industry focuses on one architecture ‘ and of course, aims to dictate the elements itself.

Motorola also set out its stall in the area of fabric, taking a contrary position to Intel. The choice of switching fabric determines which blades can be used in a system, and as in all standards, reducing the variety of options to a minimum will lead to the maximum scale, interoperability and price reductions. Therefore there is impetus to support just one fabric, but Motorola has chosen Ethernet over Intel’s favored PCIExpress.