AdvancedTCA - Evolution and beyond
After more than a decade of deployment in communications infrastructure applications, evolutionary enhancements to the AdvancedTCA (ATCA) standard are opening new market opportunities and providing the performance required for next-generation networks. The following outlines incremental enhancements that will be seen in the ATCA standard through 2020, and sheds light on PICMG's post-ATCA technology roadmap.
AdvancedTCA (ATCA) has been around for more than 10 years now, and is enjoying fairly broad adoption in its original intended marketplace, central office telecom equipment. The exact size of the established ATCA market is difficult to determine precisely – as some installed equipment is entirely supplied by ATCA vendors, some is mixed with off-the-shelf board and subsystem products supplied by the customer, and some are entirely purpose-built by telecom equipment suppliers – but it is certainly a multi-billion dollar market. ATCA is now being used in other applications, including military, where high performance and High Availability (HA) are required and commercial-grade “pizza box” technology just isn’t rugged enough.
One reason for ATCA’s success since 2001 is that it was designed to telecom requirements rather than as a purpose-built solution. For instance, ATCA is the only truly open industry standard that can provide NEBS compliance, which is still crucial and required for many applications. The ATCA standard specifies that a system operate continuously in a 55 ºC environment, which can occur when a building’s cooling infrastructure fails. Shock and vibration requirements are also higher than for standard commercial servers, and while this may be needed less in the future as network functions become virtualized, it is still an important requirement for many installations. Many that believe future network infrastructures will comprise of ever-larger interconnected networks of cheap servers do not understand these requirements.
Continuing evolution of an established standard
ATCA has undergone numerous performance improvements over the last 10 years, and these developments will continue. Two years ago the 40 Gigabit per second (Gbps) update quadrupled the backplane bandwidth of ATCA, which now approaches 10 Tbps for a single, full-mesh 16-slot system. After four years of work, an extension of ATCA known as PICMG 3.7 will be released in 2014 that makes ATCA more attractive for datacenter applications. This enhancement provides for double-deep systems with blades that plug into both the front and back sides of a rack, as well as enhanced platform management functions. PICMG 3.7 also provides a clear definition for double-wide blades that can use bigger processors and heatsinks, more and cheaper full-height DIMM memory, and more on-board mass storage. These double-wide blades can dissipate up to 800 W each when used in shelves that provide for it.
The requirements of datacenter and telecom central office will continue to blur in terms of function, but there will still be distinct minimum hardware requirements for each. Recognizing these developments, as well as trends in high-bandwidth mobility and cloud services, the PCI Industrial Computer Manufacturers Group (PICMG) will continue to evolve the ATCA standard by building on its proven track record. Figure 1 depicts an ATCA roadmap developed by a working group of key PICMG members in 2013.
The next major revision of the core ATCA specification, dubbed PICMG 3 R4.0, will recognize the dominance of Ethernet in ATCA systems and consolidate the ATCA Base Specification (PICMG 3.0), 40G Ethernet (PICMG 3.1 R2) specification, and the extensions defined in PICMG 3.7. Specific enhancements planned for PICMG 3 R4.0 are 100 Gb per channel transfer rates, which will be implemented upon the IEEE’s completion of the “100 Gb copper” Ethernet standard. Additionally, full IPv6 support will be added to the ATCA and HPM.x specifications. Further upgrades to cooling, power distribution, energy efficiency, and platform management will also be considered. Work on this is projected to begin in 2014.
The PICMG 3 R4.0 specification will enable continued construction of modular, open standards-based systems and ensure forward and backward compatibility with the installed equipment base. Market drivers for the performance improvements in PICMG 3 R4.0 include:
- Convergence of datacenter and telecom
- Compatibility with telecom cloud environments
- Network Functions Virtualization (NFV)
- Software-Defined Networking (SDN)
- Backwards-compatible capacity and density improvements
- Improved ease of use and deployment speed
Considerations for a next-generation modular platform
While PICMG and its members are committed to ATCA for the long term, at some point in the future a new platform that may not be backwards compatible with ATCA will be needed. A PICMG roadmap team is already defining requirements for this new platform, which will be called GEN4™.
GEN4 will be a new architecture that is complementary to AdvancedTCA. While the boards and shelves will almost certainly not be plug compatible, the software and management infrastructure elements will be adaptable between architectures. Though the roadmap team is focused on requirements and not mechanical or architectural specifics at this time, the GEN4 architecture will be optimized for very high-performance computation and networking applications that are beyond the capacity of ATCA. GEN4 will achieve an order of magnitude improvement over original PICMG 3.0 systems in multiple dimensions, including:
- System throughput to hundreds of Tbps, module bandwidth to tens of Tbps
- Physical module volume of 5-10 L designed for off-the-shelf silicon options
- Module cooling capacity over 2000 W with fluid cooling options
- Efficient power delivery with high-voltage DC options
- Scalability to efficiently create large multi-frame systems
- Improved Size, Weight, and Power (SWaP)
- Integration of optical interconnects where appropriate
- Reduced CAPEX and OPEX, including rapid service deployment capability and reduced integration expense
With the explosive growth of data traffic driven by big data, the Internet of Things (IoT), and WEB 3.0, GEN4 will be positioned as the standard modular platform architecture of choice for High-Performance Computing (HPC), central network, network edge, and high-capacity storage applications from 2015 through at least 2025. The development of GEN4 is intended to provide a new approach to varying performance needs through a hierarchy of increasing capability subsystems that support SDN and NFV, while providing investment protection for customers with deployed AdvancedTCA systems. Over the next few months, PICMG will begin to publish preliminary roadmaps, detailed requirements, and architectural elements for the development of the GEN4 platform (Figure 2).
An active role in the future
PICMG is welcoming system architects, telecom service providers, datacenter designers, and potential users to join the effort to fully define GEN4 and produce a working standard in a sensible timeframe. If you are interested in contributing to this effort and becoming a PICMG member, please contact PICMG at [email protected].