In a November 4th, 2009 e-cast on "Open Standards Based Platforms Ideal for Military/Aerospace/and Government Applications", sponsored by the Intel® Embedded Alliance, the issues around standards-based platforms using long-life Intel embedded processors were explored by a panel of experts.
Chris Ciufo, editorial director of Military Embedded Systems, moderated the event for OpenSystems Media. Panelists presenting were:
- Peter Carlston, Platform Architect, Intel Embedded Computing Division
- James Doyle, Congressional Affairs Liaison, Emerson Network Power (with Rob Persons, Field Applications Engineer, during Q&A)
- John Long, Product Line Manager, RadiSys Corporation (with Lorraine Orcino, Product Marketing Manager, during Q&A)
Chris opened the e-cast with brief comments about the network centric warfare vision of "MAG" applications requiring interoperability between both legacy systems and new systems with advanced features. One way to achieve that is using open standards based, Commercial Off the Shelf (COTS) products. Chris then posed the guiding question for the event: Can you really meet strict requirements and deliver better products faster, and at a lower cost, using open standards? The short answer is "yes", but the panel was asked to prove that.
Peter Carlston then presented a couple of key ideas, starting with the idea of COTS products operating in one network - a really large network, with a lot of pieces, but nonetheless one network. The idea of long-term availability is important, with all products on the Intel embedded roadmap having 7 to 10 year availability. He also mentioned hardware virtualization and secure boot enhancements, features also supported in Intel Architecture processors. Peter cited Nehalem as an example of a cutting-edge processor targeting the types of signal processing algorithms often found in MAG applications, and other Intel Architecture processors fitting lower TDP points with good performance.
James Doyle commented on the fit of open systems platforms in C4ISR, networking, and situational awareness applications for the warfighter. He cited a quite example of the potential: "Recently in Afghanistan, there was a situation where the first person on a scene literally wrote a blog in his HMMWV. Within 7 minutes, that information was directly in front of the President, who was able to make a decision and act on that intelligence." He positioned SWaP - Size, Weight, and Power - as an important criteria supported by open standards platforms. Standards he mentioned prominently were AdvancedTCA and VME.
John Long commented on the mapping of COM Express to smaller, handheld applications, while AdvancedTCA maps well to larger C4ISR applications requiring higher performance. AdvancedTCA offers the high performance of larger Intel Architecture processors combined with 40GbE and other high performance I/O. COM Express fits a mobile warfighter need, in small, battery powered designs, or in small platforms like UAVs.
After these overviews, the Q&A portion of the e-cast began with extended panel discussion of questions submitted by attendees. Following are excerpts from that session.
For more information:
Register to view the archived event, with slides and audio: http://bit.ly/34DZx8
Download the podcast with audio only: http://bit.ly/3Ij9ZC
CC: Why is Intel's embedded group focusing on the military again? Didn't you get out of the mil-spec business?
PC: Yes, we did, but with the renewed emphasis on commercial off the shelf silicon and systems built with that silicon by our partners such as Emerson and RadiSys, we decided the time was right to focus again on the military. We now have very power efficient processors which can go into a large number of form factors including small and very small form factors, and we're getting a lot of traction .... We believe we have the right product, the right support mechanisms, and we're able to enable our partners to be successful in MAG applications.
JD: We are seeing a growing change in prime contractors and the DoD with interest in Intel silicon. As things move forward, we see greater adoption for commercially available silicon that operates in all different temps and in different platforms [for these MAG applications]. We think it's a great time for Intel to come back to this marketplace.
JL: We've also seen the adoption of Intel Architecture processors in MAG applications. They have lower power, high performance processors, and also higher performance parts. This allows a lot of applications to move to more standard form factors, and also we're seeing movement to more standard operating systems.
CC: 7 to 10 years historically hasn't been enough support time. My opinion is ... this is what we get with COTS; a component supplier offering a 7 year life cycle is as good as you'll get from any component supplier with only a handful of exceptions. What do you think?
PC: I agree. There are many military programs that equipment is being sold into for much longer than 7 to 10 years. That's where a commercial company such as Intel has to rely on our partners to manage the supply chain. We give along early advance notice of last time buys to help our partners.
CC: What is the future of warfare, and how do you see the embedded computing industry supporting it?
JD: Let me try to wrap this around situational awareness. Our view is the collaboration of technology and the warfighter will continually enhance our capability to be able to operate in multiple theaters, not just US forces but forces around the world. Being able to use COTS technology allows us to collaborate with our partners, to be able to share information to do different missions ... We see technology at the forefront of allowing us to have a "more aware" battlefield which will keep our troops safer.
RP: I agree with you in general. I think that having a variety of form factors, not just traditional VME, fits a variety of applications. Because we have all these different open standards all using similar silicon, I think we can address a variety of different applications but have a lot more overlap in software.
CC: Does Emerson support any other open standards besides AdvancedTCA?
RP: Our entire product portfolio really centers around open standards. This includes AdvancedTCA, VME, CompactPCI, and even some newer ones including AMC and MicroTCA.
CC: For RadiSys, where can COM Express and AdvancedTCA be used in the MAG market?
JL: There are two areas. COM Express is really good for mobile applications. We've seen adoption in computer display units where you can easily pull them out of a vehicle, in the man-wearable devices, and in smaller UAVs. AdvancedTCA is a much larger form factor and is probably not viable for a tank or fighter plane, but in a large aircraft and ground stations it's ideal. When we look at all the different form factors out there, they all fit in different segments and are all viable in the long run. Some are better for different applications than others.
CC: For Intel, what is Intel doing to address the need for much higher performance with the smallest SWaP envelope possible?
PC: Intel as a corporation is very focused on SWaP. It's no secret that in the past there was a need for more and more performance, and in the desktop space you could just add bigger fans and more ducting and that sort thing. Those days are over. With the move to multicore processors we've really enhanced our technology such as fine-grain power management where we shut off parts of the core and parts of the "uncore" when they're not being used, and microseconds or even nanoseconds ahead of when they need to be used they're switched back on. There's a lot of technology going into making our processors more power efficient ... In the Nehalem microarchitecture, we've quadrupled memory bandwidth with memory access times down to 4 or 5 nanoseconds, and that's made a big difference in the performance of a number of algorithms. In a high end radar algorithm, between an unthreaded application and a multithreaded application we've increased performance 38 times by running on a new Nehalem family processor.
CC: For Emerson, what are the chief concerns being raised that inhibit the adoption of COTS technology in MAG applications?
JD: One of the concerns is certainly lifecycle. Understanding the supply chain management and how we continue to support these programs for 15 to 25 years depending on the program is important. The other areas for COTS really come down to Congress, and understanding the acquisition cycle, and the way you need to get to the left of an RFP, and help the customers and the users understand what is available. I ask people to write their congressman, and ask them to change one phrase - please make sure that in next resolution it says "shall be open standard" instead of "should be open standard".
CC: For RadiSys, how important are middleware standards for high availability, such as SAF, to the success of AdvancedTCA in MAG?
JL: Middleware is critical and open standards are very important. Open standards provide some vendor independence and allow taking the same base product and moving it to multiple applications. Open standards build on each other get you cost effective COTS products.
CC: Would anyone like to comment on VITA's new OpenVPX initiative and how it addresses high availability?
RP: OpenVPX is definitely going to benefit from the work already done in the AdvancedTCA world, but the systems will be treated differently. We'll unlikely see hot swapping capabilities in OpenVPX, but having an internal view of the system and being able to provision a variety of different boards within a chassis will be very strong. At least a subset of the OpenSAF architectures will bleed into OpenVPX systems management.
JL: I'd agree, the use of the same management strategy is good. It allows reuse of technology. One of the things I hope to see in the long run is this will allow folks to easily scale from VPX to AdvancedTCA by reusing a lot of stuff already developed.
CC: Do you think that's a likely scenario? That someone would scale from a 3U VPX system to a rackmounted AdvancedTCA implementation?
JL: I look at large companies addressing different applications; a lot of the base middleware and technology from the software perspective can be reused across different platforms. That's where the value of the SAF standards is.
RP: I agree, and I think that it's very much like you see with Linux deployed in a lot more areas - having that common software thread that can be reused, whether in a hardened environment or not-so-hardened. The same set of software can be used in a variety of different platform styles.
CC: For Intel, how does virtualization affect the MAG market?
PC: There is a lot of interest in virtualization, starting with information assurance platforms. These are platforms where you have to protect vital information from software-based attacks, viruses, root kits, all that sort of thing. It's also about running multiple levels of security at the same time on the same platform. Intel's virtualization technology implemented in our processors allows building very high performance, multiple levels of security. We have other technology that aids in secure booting. In other types of applications, like shipboard computing, reducing racks of equipment with more power efficient servers - more like a data center - is a model for virtualization.
CC: For Emerson, can technology in COTS systems and applications be leveraged and reused across multiple systems and form factors without sacrificing security?
JD: Security is an abstract of an organization, its technology, its management, its value chain from beginning to end, including people. It comes down to people, process, and the way technology is utilized. A president of a major prime contractor said: "I'm convinced that we have six non-national spies within our organization, today." Depending on how you look at your organization, how you manage it, and how you deploy systems, security is a critical aspect.
RP: We're seeing a lot of the software that runs on these systems taking advantage of closing a lot of those security gaps. It's a concerted effort between how you deploy systems, and who you allow to access them, along with some of the underlying software technology that will be a very strong solution.
CC: For RadiSys, is AdvancedTCA viable for airborne operations?
JL: Yes, it is, especially in larger aircraft. There are customers deploying AdvancedTCA in applications such as surveillance. It is very viable and the challenge in airborne is hardening boards enough to handle high frequency vibration and any unique cooling requirements.
CC: For Intel, is multiprocessor debugging technology progressing with multicore use in embedded systems?
PC: Obviously, you can't just keep the application the same and rely on processors running ever faster. There's a lot of work going on to enable customers; there's not very much understanding of multithreading, there's not a lot of experience with it , and yet the performance requirements are just not going to be met without a lot of multithreading going on. Intel is putting a lot of effort into our multithreading tools. We have thread checking tools, we have debugging tools, we have a lot of things running on Linux, and we're working real-time operating systems vendors to improve their tools. It's happening, but maybe not as fast as I'd like to see just yet - that's just my opinion.
RP: In the case of utilization, that's where virtualization will become very important. We see it especially in the telecom space where some of these AdvancedTCA blades can have up to eight, heading toward twelve cores - being able to use all those cores is going to be very difficult without virtualization. It's going to be important to take advantage of all the power we're producing on a board.
CC: For RadiSys, an audience member asks: "We've had interest in a 2 slot AMC system for the military for mobile applications. How could COM Express compete with the number of AMCs already out there?"
LO: The selection of COM Express versus AMC is driven by the application requirements. We are seeing deployments with COM Express in these mobile MAG applications.
CC: As a follow on, how does VPX compare to AdvancedTCA?
JL: We're excited that VPX has been adopted as a standard, and we think it fits a good niche in the mission computing role. The differences? AdvancedTCA supplies more input power, higher processing compute power, and has been around longer - a lot of the bugs have been worked out. There are applications where VPX is a better fit.
JD: The big difference from our view is really the deployment. AdvancedTCA is great for "benign" systems - it's a large platform that is designed for communication applications. OpenVPX is for much more rugged environments, in 3U and 6U form factors.
CC: For Intel, which OS will dominate - Windows, Linux, VxWorks, or others?
PC: Well, it's all speculation. We're seeing increasing adoption of Linux, even in some surprising applications that previously have used an RTOS. On the other hand, there are certain applications that are going to need a true real-time operating system forever. I think it's going to be a mix. There's a lot of work going on for small devices with things like Moblin.
CC: For Intel, with shrinking geometries cause more errors due to single event upsets (radiation)?
PC: It's a question I get all the time. Some of the literature says this is going to happen, while other says there are some mitigating circumstances going on. As geometries get smaller, the jury is still out. We are paying a lot of attention in our packaging, our material science, to mitigate these kinds of effects.
CC: For Intel, what products are currently available to support high throughput signal processing?
PC: With our dual socket and quad socket processors on bladed form factors, we support FPGA plug-in modules that attach to the processor bus - tightly coupling the FPGA with the general purpose CPU. They share the same memory, they have cache coherency, all that. We've licensed our RTL to both Altera and Xilinx for this purpose. Another part of this is what we call the acceleration abstraction layer, which allows abstracting those hardware resources in a service oriented architecture layer so the system can publish what's available and the application can take advantage of the best acceleration technology that it needs to at that moment. Also, as our customers move from Freescale AltiVec to Intel SSE, we have tools available to automate that conversion.
CC: Really? Tools available to cross compile from AltiVec to SSE?
PC: For the digital signal and image processing portions, yes. For example, there's one tool that takes VSIPL libraries for Intel Architecture and performs very well, and there's another one that has an altivec.h header file for Intel Architecture which can be linked in. Both of these take care of things like byte swapping.
CC: For Intel, what types of feedback are you getting from the MAG market?
PC: There's a lot around our graphics processor called Larrabee. I can't discuss details or schedules ... What we're seeing are very difficult compute problems from our customers. GPUs are getting incredibly capable, with huge amounts of data coming in. Many core processors, 80 to 100 cores in our research, are looking at some of these very high end workloads. It'll be more of a hybrid system, putting CPUs and GPUs together to get the best solutions.
From the Q&A session, you can see that the interest in this application space is rather intense, and the innovation coming from Intel and the Intel Embedded Alliance is beginning to produce ideas that are shaping the future direction of military compute applications.
What are your thoughts on the ideas discussed during this event? What do you see as the driving issues for MAG applications and how they are being approached? What would you like to see next on this topic?