No Resting On Laurels for ATCA: Q&A with PICMG’s Joe Pavlat
Sometimes more is more. The efforts PICMG is making on behalf of ATCA and other embedded standards include more board real estate (PICMG spec 3.7), more speed (100 Gbps simulations proceed apace) and even more confidence in the part standards organizations play in creating meaningful specifications.
Editor’s note: It was after an informative interview with my friend Joe Pavlat, president and chairman of one of the primary open standards organizations, PICMG, that I wrote PICMG Picks Up Steam–All the Way to the Red Planet to catch readers up on the evolution of the COM Express, CompactPCI and ATCA specs. The backdrop to my most recent talk with Joe includes—among much else—the Internet’s march toward heterogeneous local intelligence, faster Ethernet, and attempts to topple Intel server space hegemony. Edited excerpts follow:
Chris (C2) Ciufo: Joe, what’s the latest on faster Ethernet (100 Gbps), ATCA and COM Express.
Joe Pavlat: Work on speeding up AdvancedTCA to provide 100G Ethernet signaling is moving forwards and the spec is due to be released by the end of 2015. The simulations are going well. PICMG has long believed multi-vendor interoperability is important, whereby a user can buy different bits and pieces of a system from different vendors and when put together, everything works. The IEEE, which controls the Ethernet standard, didn’t provide any guidance on allocating budgets for things like cross talk and insertion losses to the various boards, backplanes and the connectors the signals must travel through. We are paying careful attention to this and think we have it right.
PICMG also recently released PICMG 3.7, or AdvancedTCA Extensions. This spec is a major enhancement to ATCA. It fully defines double-wide boards capable of dissipating up to 800W each. It also defines double-deep systems where two systems are interconnected in a front-to-back fashion. This dramatically increases board real estate and minimizes the amount of valuable vertical real estate in a rack. PICMG 3.7 also provides a great deal of thermal design information and is, I believe, the best thermal design guidebook for rack-based systems there is.
Not a lot is going on development-wise in the very small form factors like COM Express, although at last year’s Embedded World trade show in Nuremburg the general consensus of the manufacturers I talked to is that about 2 million COM Express units are shipping per year. That number is expected to increase. The increasingly popular CompactPCI Serial standard is just about to be released in a new version that provides more flexibility in implementing board-to-board Ethernet capability without the need for a switch. It also better defines rear panel I/O, which is important for critical conduction-cooled applications.
C2: We can’t go too far these days without reading about Network Functions Virtualization (NFV) and Software Defined Networking (SDN), whether from an Open Plane (S/W) standpoint or the evolution of the Internet, or ways in which customers are trying to break the lock of the equipment providers. Give us your opinion on these.
Joe: Software Defined Networking and Network Functions Virtualization are important tools that can allow customers to rapidly add features to their networks or to reconfigure them without upgrading a lot of hardware. These tools are useful in some places and not in others. There are really two networking worlds out there, and they are very different. The Communication Service Providers (CSPs) like Verizon, AT&T, Comcast, etc., have a very different set of requirements than the Enterprise Service Providers (ESPs) such as Amazon, Facebook and Rackspace. The CSPs must provide data across their large networks with little latency—think mobile video—and they have reliability requirements that are much higher than the ESPs. When you post a message to Facebook, it doesn’t really matter if it takes 1 millisecond or a hundred to transfer a packet of information. In the CSP world, where voice and video must be delivered seamlessly, latency and reliability are critically important. For some functions like 911 services, reliability and “up time” are mandated by law. It is a common misconception that high-reliability telecom networks can be built with cheap commodity servers like the type the ESPs use. In fact, Verizon and AT&T claim that only 10% of their future equipment requirements can be met with commodity “pizza boxes.”
C2: A presentation from the Open Server Summit openly predicts that standards such as ATCA are on their way out: too expensive, too much/too late, and just not fast enough technological evolution to keep pace with the market. Your thoughts?
Joe: That presentation’s conclusion notes the need for an ecosystem of COTS hardware tuned to meet the needs of the communications service providers (rather than enterprise) [See above]. PICMG and other standards organizations play a key role in making sure that multivendor interoperability will be maintained in this emerging ecosystem.
ATCA is a long way from being on the way out. It remains the most popular open hardware platform used in telecom today. What is changing is the software that runs on it. SDN and NFV allow more flexibility without replacing or adding hardware. PICMG began work about two years ago to define a next-generation platform that was bigger, faster, more extensible and generally meaner than ATCA. We talked to a number of CSPs, and they indicated their highest priority now isn’t new hardware per se, but rather increasing their networks throughput. Video, which is expected to be over 80% of telecom network traffic in a few years, is a data hog when compared to simple voice communications. Verizon told us that it needs to increase its network throughput 60% per year for the foreseeable future. That is a staggering 10x increase in just five years and 100x in 10 years. So PICMG is focusing on improving the existing platforms like ATCA through technologies like video acceleration and 100G data transfer. At 40G, the current top speed specified for ATCA, a full mesh, 16-slot system is capable of handling just over 10 terabits of information per second. As I mentioned earlier, high availability and high system reliability are proven features of ATCA, and there aren’t any other open standards out there that can provide that.
C2: I’ve recently noted how the Internet’s evolving from big core/central servers to disaggregated and heterogeneous local intelligence and processing. From Intelligent Gateways (Intel’s term) to CPU + GPGPU (AMD’s APUs) to ARM’s latest A72 to Qualcomm’s ante up…. the Internet’s data-packet-centric architecture’s changing. What are the ramifications to standards groups like PICMG?
Joe: The Internet is a big place with many technologies and players. The trends that you mention are only a few of the explorations that are going on as operators strive to reduce expenses and provide new revenue streams and services.
The need for standards organizations does not diminish with the acceleration of innovation. To the contrary, industry specifications promote multivendor interoperability and help prevent ecosystem fragmentation. PICMG’s role as a standards organization is to bring industry players together in a safe harbor where meaningful specifications are developed.
C2: Is PICMG’s day in the sun over?
Joe: Quite the opposite, in fact. PICMG has been active and producing specifications wanted by our members for over 20 years. Unlike many of the “one trick pony” standards organizations that pop up, release one or two things, and then disappear, PICMG is stable. Our processes are well proven and our intellectual property policies are strong. It is true we are putting most of our efforts on new revisions of existing standards right now and not working much on very new and exciting stuff. That’s a cycle that changes every few years.
Chris A. Ciufo is editor-in-chief for embedded content at Extension Media, which includes the EECatalog print and digital publications and website, Embedded Intel® Solutions, and other related blogs and embedded channels. He has 29 years of embedded technology experience, and has degrees in electrical engineering, and in materials science, emphasizing solid-state physics. He can be reached at email@example.com.