AdvancedTCA (ATCA) continues to evolve to meet not just the market demands for the telecom central office, but networking, data center, medical, and military communications applications. The two main goals that ATCA suppliers have been trying to solve often conflict with each other: offering more performance in less space, while meeting the specific market and compliance requirements for each industry.
In the early days of ATCA, the architecture had a viable ecosystem for central office applications using vertical 12U-13U shelves in 14 slots. For the central office, the shelves required full redundancy and FRU (field replaceable units), 48V DC power feeds, and 200W per slot front-torear cooling performance. Thus many 12U sizes were later abandoned as cooling demands increased.
Not long after, horizontal mount chassis entered the market in 4U-6U heights for telecom and networking applications with backplanes in 5 or 6 slots. AC power options were often a key requirement for these applications (and therefore units like the 4U, where space is limited for AC power, shelf managers, etc., also largely faded away). But, these early solutions had side-to-side cooling and were limited for telecom applications that require NEBS-compliance, and virtually ignored other applications such as military and ruggedized applications. Further, as chip advances continue, providers need to find practical solutions to cooling, signal integrity, and EMI. Plus, cooling demands have changed the focus from shelves per rack to overall performance density.
There have been two main pushes for performance density. One of these efforts is to increase the signal speeds. The move to four lanes of 3.125 Gbps into a single lane (port) for 10GBASEKR and KX4 and bundling four 10 Gbps lanes into one 40 Gbps channel (40GBase-KR4) has been addressed by the IEEE 802.3b/ba subcommittees. At 40 Gbps speeds, there is a significant performance gain across the shelf. Additionally, simply going from Dual Star to Full Mesh designs can add significant bandwidth. Before we go to 100G efforts, there is certainly opportunity to incorporate Full Mesh across 40G for even more performance gains.
The second approach is to offer more performance using less chassis space by increasing the computing density at the blade level. Carrier grade solutions have gone from 12U to 13U and up to 15U for the additional cooling required for 350W+ per slot of today’s ATCA boards. Plus, the RTM section can add another 40-70W per slot of heat to dissipate. But, with 40G options, the bandwidth/performance density is improved off the bat. With the push to 40G, conducted and radiated EMI immunity needed to be changed, posing further challenges to the increased cooling requirements. Many 5U shelves have moved to 6U to increase cooling, but also offer the fully redundant AC power modules, shelf managers, etc. AC power is a key requirement for the flexibility of use in a wide range of applications outside of the central office. It is more conducive to medical, data center, networking, and some mil/aero applications (where a combined AC/DC solution is attractive).
Aside from the signal speeds, there is a way to increase ATCA computing density by 50% in a horizontal 6-slot chassis. By integrating the switch fabric functionality into redundant shelf managers, two slots that are usually dedicated switch fabric slots in the ATCA shelf are saved. The two extra slots are utilized as standard payload (node) slots, which greatly increases the performance density. Thus, instead of having only 4 payload slots in the 6U shelf, there are a full 6 slots available.