Developers needing high-performance, high-reliability computing systems have new design options becoming available. The PCI Industrial Computer Manufacturers' Group (PICMG) is now expanding its Advanced Telecommunications Computing Architecture (ATCA) specification to support additional applications beyond the telecom industry.
In particular, fresh options such as double-height blades, double-sided chassis entry, and significantly higher power limits promise to expand the applicability of ATCA solutions.
This expansion combined with ATCA's proven development ecosystem will bring the cost and logistics benefits of high-reliability modular computing to a wide variety of mission critical applications in markets such as the data center and military/aerospace.
While the ATCA specifications target telecom applications, they have characteristics that make the architecture of potential interest to other industries.
To be compliant with telecom's NEBS (Network Equipment Building System) standards, for instance, ATCA had to provide hardware and software mechanisms for implementation of high-availability and fault-tolerant design.
It also had to handle specified environmental conditions, including operation in 55?C ambient temperatures, shock and vibration corresponding to severe earthquakes, and the like. The specifications are now in the process of expanding to include greater power and spacing options, along with higher chassis densities to further accommodate needs beyond telecom.
The design of ATCA hardware gives it the flexibility to be applicable outside telecom. The architecture revolves around a modular blade computing structure (Figure 1, below ) with blades supporting hot-swap capability for easy system maintenance.
|Figure 1 ” ATCA blades support several different sizes of AMC modules, giving designers considerable flexibility in customizing blade operation using off-the-shelf equipment.|
Individual blades provide modularity by supporting the presence of Advanced Mezzanine Card (AMC) modules that can hold computing elements, I/O (input / output) interfaces, and other high-level functionality. Developers can use a standard module frame and customize it for their specific requirements simply by populating it with the appropriate AMC modules.
The AMC modules are large enough to serve as platforms for complete computing subsystems of considerable complexity and performance. Four different size modules have been defined that will mate with a standard ATCA blade.
These different possibilities allow module designers as well as system developers the flexibility to utilize only as much space as the function requires, thereby saving cost. Furthermore, developers can use AMC modules natively within a compact MicroTCA chassis to build low-profile systems with powerful computing, storage and I/O capabilities.
Switched Backplane Combines Performance & Flexibility
ATCA blades connect together over a switched serial backplane with a high data throughput. The backplane supports as many as 21 serial channels to each blade, and each channel can handle data at rates up to 12 Gbps ” providing a combined data throughput capacity of 2.4 Tbps for an ATCA shelf.
The serial channels are protocol-agnostic, freeing developers to utilize the serial communications standard of their choice. Possible candidates include but are not limited to 10 Gigabit Ethernet, Serial RapidIO, and PCI Express.
The ATCA specifications go beyond hardware to include many key software elements for the creation of a high availability system. They include a chassis management system, for instance, that provides for module-level control, fault detection, and system power management (Figure 2 below ).
|Figure 2. The ATCA specifications include details on a full module and system management structure that supports capabilities such as automatic fault detection and correction as well as hot-swap system maintenance.|
The management system uses IPMI (Integrated Peripheral Management Interface) signaling and a central controller to query the status of AMC modules on the blades and to control their access to the backplane.
This gives the system a mechanism for detecting failures and isolating failed modules from the system to prevent system crashes. It also supports hot-swap operation by shutting down modules during the change-over.
Other standardized software elements of ATCA leverage this chassis management system to implement high availability and fault-tolerant system operation. A Carrier Grade Linux operating system provides the foundation for this system software.
High availability middleware provides the mechanisms and programming interface for enforcing fault tolerance. Developers only need to add their application software to complete the design of a high availability system.
Supporting the xTCA Specifications
To ensure that original equipment manufacturers (OEMs) are able to fulfill the promise of creating a fully-functional telecom system that meets their needs from commercial off-the-shelf (COTS) components, an entire vendor support ecosystem has arisen around xTCA (consisting of ATCA, MicroTCA and AMCs). PICMG maintains the hardware specifications, resolving errors and adding new capabilities as needed.
The Service Availability Forum (SAF) maintains the software specifications, including the high availability middleware. Multiple vendors have developed modules, blades, chassis, software and other system elements in accordance with the specifications, giving system developers a wide choice of suppliers.
The Communications Platforms Trade Association (CP-TA) has developed standardized tests and procedures as well as certification guidelines for ensuring that these multiple sources have maximum interoperability.
The benefits of the standardized modular design approach are compelling. It is highly configurable, for example, which allows developers with a diverse range of requirements to meet their unique needs by mixing and matching standard components. It is also highly upgradeable. System developers simply change an AMC module in order to incorporate the latest technology into their design.
The multi-vendor supply environment provides several key benefits due to competition among vendors. One is innovation: suppliers will work to provide unique benefits in order to differentiate themselves. At the same time, the basis in a standard ensures that multiple vendors offer comparable products so that developers are not tied to a single source of supply.
The approach also offers developers significant cost benefits. The ability to use COTS system elements considerably reduces system design effort. Meanwhile, competition helps keep component prices down. Further, vendors of standards-based products have access to a wider customer base than for custom-designed products, providing manufacturing economies of scale that can be passed along to the system developer.
xTCA Broadens Appeal Beyond Telecom
While xTCA and the ecosystem that supports it initially targeted telecom applications, the power and flexibility that the architecture offers has the potential for much broader appeal. Financial computing systems, such as for banking and investment management, are one possibility.
These systems handle continent-wide and world-wide transactions, creating a need for high data throughput over a mix of serial protocols. Financial systems also need fault tolerance in order to avoid loss of information if a failure occurs in the middle of a transaction.
Enterprise data centers are another candidate for the xTCA architecture, where a critical need is for an expandable fabric that can grow with the business. The ability to readily upgrade the system is also important, since it avoids the disruption of switching over to a new system as technology evolves. These two attributes are also key for video processing systems such as video-on-demand servers, which can benefit from on-the-fly reconfigurability of connections through the switched serial fabric.
Not least among potential candidates for xTCA are military systems. xTCA's fault tolerance and capacity for hot-swap maintenance provide the kinds of high availability required in combat support systems. Further, PICMG recently released specifications that extend the range of environmental conditions that xTCA hardware ” specifically ruggedized AMC modules ” can be made to tolerate. The new specification includes temperature extensions as well as additional support for vibration and shock resistant design.
Recently PICMG has begun development of even more extensions to the xTCA specifications, with an eye on simplifying the application of xTCA components to these candidates. One such extension is the development of a double-height ATCA blade. In the original specification, blades all occupied the same space in an ATCA chassis ” a pitch of 1.2 inches.
The double-height extension to the specification would permit chassis and blade designs supporting a wider pitch, which will give designers a larger envelope to work within for incorporating bulky items such as hard disk drives and heat sinks with integral cooling fans.
Double-height also allows for greater air volume and flow between boards, increasing cooling capacity. This increased cooling, in turn, allows designers to add more high-heat components, such as larger and denser memories, to the blades.
Power and Chassis Design Options
Increased power capacity is another ATCA extension that PICMG is developing. The original ATCA specification called for power of 200W per blade. Many of the candidate applications, however, have a need for processors and memory densities that will combine to exceed that limit. How much more power would be practical is not yet clear, though, so the magnitude of the power extension that PICMG will allow has not been settled.
Another extension to the ATCA specification under consideration is the allowance of double-sided chassis designs. In such systems blades could be plugged into a common backplane from both the front and the rear of the chassis. This helps increase the vertical compute density a chassis can achieve, helping high-capacity systems fit into smaller spaces such as equipment closets in office buildings.
All of these extensions will have a common requirement of backward compatibility with existing blades and modules. Those currently on the market and designed to the original specifications must be able to also work in a system designed to the extended specifications.
One of the key attributes of the xTCA design approach is that component vendors can serve a broad market with a given module design. If the ATCA extensions were to not require backward compatibility, that attribute and its benefits would be compromised.
As with any new specifications, these xTCA extensions have the potential of being ambiguous or undefined in some areas of detail, leading to multiple interpretations and incompatible implementations.
Yet assured multi-vendor interoperability is an essential element of standards-based system design. Indeed, it is the key to mix-and-match design and easy upgradeability and helps eliminate system integration conflicts that could wipe out any savings in initial system design time.
Assuring interoperability is the role of CP-TA in the xTCA ecosystem. The organization has defined test tools, test methods, and reporting standards for evaluating interoperability in several key aspects of the xTCA specifications. Further, the organization has developed interoperability compliance levels and corresponding certification guidelines that give system developers direction in evaluating how much integration effort a given product choice may entail.
Currently, CP-TA has focused its efforts on the areas of thermal, mechanical, and module management compatibility. In thermal compatibility, for instance, the CP-TA testing and compliance levels help developers to avoid creating system hot spots that could otherwise arise if adjacent modules from different vendors had high-power dissipation elements in proximity.
Similarly, they help developers avoid creating airflow restrictions because of component height and placement on modules. CP-TA programs also help ensure that a module's specific implementation of the system management specifications does not generate integration conflicts with the chassis controller or other modules.
CP-TA will, in collaboration with other industry organizations, continue to expand the scope of its compatibility assurance efforts to other aspects of xTCA specifications.
This expansion includes working with PICMG to understand the potential for compatibility issues arising from the ATCA extensions, and ensuring that the needs of these new applications areas are adequately addressed. Vendors and system developers alike share a common interest in seeing that the benefits of modular design are realized through assured interoperability, and CP-TA welcomes the involvement of all such interested parties.
The extensions to ATCA that PICMG is developing will open the architecture to new application opportunities. Developers in these application spaces should explore the architecture and evaluate the benefits that it can provide. Then, together with organizations such as CP-TA, they should join in the decision-making process to help ensure that those benefits will remain intact as the specifications evolve.
Sven Freudenfeld is a Director and President of the Communications Platforms Trade Association as well as responsible for North American Business Development for the Kontron AG line of AdvancedTCA, AdvancedMC, MicroTCA, and Pre-Integrated Solutions. Sven earned his electrical engineering degree in Germany, and is Chair of the CP-TA marketing workgroup focusing on the interoperability of COTS standard building blocks.