The modern battlefield is seeing dramatic increases in the volume of data generated, fueled by continued growth in the number and capabilities of sensors and automated systems. For example, high-definition video cameras can generate 8 to 10 Mbps of streaming data per video feed, with multiple secure feeds required for applications such as mapping, secure chat, and augmented reality, not to mention typical ISR applications. Data sharing has become an essential weapon, enabling real-time situational awareness and its resulting competitive warfighting advantage. The scope of shared data on the battlefield is driving designers to consider strategic tech refresh moving from VME to VPX, particularly now that the VPX ecosystem is defined for board sizes, connectors, and signaling for the next 10 to 20 years. This can be a challenging migration with a dramatic shift from VME communication protocols. Designers must completely understand the VPX architecture, specifically with regard to how tech refresh transitions can be streamlined with tools to simplify data flow management in High Performance Embedded Computing (HPEC) applications. Modern Application Program Interfaces (APIs) can ease such migrations.
Strategically assessing tech refresh opportunities
Simplifying migration is essential for efficient tech refresh opportunities so that applications can capitalize on increased bandwidth and processor advances. The need for advanced connectivity and bandwidth is constantly evolving because of requirements for higher fidelity of data (higher resolutions), different types of data (video versus audio), and the amount of data sources (higher numbers of sensors). Yet, in an integrated battlefield, it is important that systems be refreshed with minimum risk to the overall system or other applications within the network. Migrating and enhancing VME-based systems to VPX meets growing performance requirements to handle, for example, multiple video and audio channels under surveillance. At the same time, sophisticated data sharing capabilities are required to allow secure processing across multiple nodes, which is often required because of the rapid increase in sensors deployed in military systems. As sensor data management increases exponentially across the battlefield, VPX-based HPEC platforms are becoming a critical resource based on their ability to integrate a range of high-performance COTS products to achieve high throughput and processing.
VPX allows board computers to move away from decades of parallel bus architectures on the backplane, with a little help from 1 GbE, and implement high-speed serial link point-to-point connections between boards. With VPX connectors and backplanes, multi-gigahertz signals warrant systems where the full dataplane bandwidth is no longer shared between boards. For example, each board can have one or more dedicated 10 Gb connections via Ethernet or PCI Express. HPEC is the first domain to benefit from the tenfold increase in I/O bandwidth between computing boards that VPX offers (Figure 1). This enables a new breed of unparalleled applications for sensor data processing platforms used in radar, sonar, and general imaging. But now, as hardware guidelines are set, OEMs and developers have to search for the ideal communication protocols. With all the advantages of current standards such as PCI Express, GbE, Serial RapidIO, and many others that can be used for intrasystem communications, this is not an easy task. The challenge for OEMs is to find an easy-to-use, yet fast and low latency communication protocol. Amongst serial link technologies available today, PCIe is poised to bring many military applications to a new level, optimum for tech refresh migrations that must be put in place with minimum risk to the overall system or application.
Streamlining migration and costs with PCIe, APIs
PCIe is a computer expansion card standard initially designed to replace the older PCI and PCI-X standards. The PCIe topology is based on point-to-point serial links, instead of a systemwide shared parallel bus architecture. This represents an ideal link between I/Os and processor units, as well as a native communications link in a multiprocessor environment. PCIe technology has evolved recently from Gen1 to Gen2, doubling the clock speed from 2.5 GHz to 5 GHz and also the theoretical throughput. And PCIe Gen3 is on the horizon, which will again double throughput. At the software level, PCIe preserves compatibility even if new capabilities such as power management or advanced error reporting are added. Designers are learning that APIs as data flow management tools can greatly simplify migration from VME to VPX, accelerating the design process.
APIs provide a thin layer of software that allows faster application development for IP-based transport over PCI Express. For example, using Kontron’s VXFabric, a PCI Express switch fabric technology for HPEC environments, OEMs can deploy high-performance 6U OpenVPX technologies with >10Gbps board-to-board connectivity while facilitating the integration of next-generation processor architectures (Figure 2). Using the plug-and-play capability of the PCI fast link, the switch fabric moves data at an ultra-high speed. The key advantage is PCIe’s performance as a native data bus in all modern processor chipsets – resulting in an expansive PCIe-based software ecosystem with broad support for peripheral interconnects.
APIs address data flow complexities
In migrating from VME to VPX, changes affect the backplane and all system cards, presenting a dramatically different scenario than a simple CPU card upgrade. This type of refresh routinely requires greater design expertise, replacing the bus-based architecture with a network-based protocol and requiring a significant retooling of application software.
API tools enable OEMs to implement efficient inter-board communication at hardware processing speeds, achieving the highest bandwidth by leveraging PCI Express (Figure 2). From the hardware point of view, the architecture is based on several CPU boards, each featuring several processing cores, interconnected through PCIe via the VPX backplane, using a PCIe switch. From a software perspective, the API is equivalent to an Ethernet network infrastructure. The API implements layers that accommodate communication with an IP socket programmatic interface, allowing direct access of classic protocols such as TCP or UDP. Development and migration efforts are streamlined as the API requires no modification of existing applications.
The standard user programming model of an API tool is based on the IP protocol and implements a socket layer API through an emulation of an Ethernet interface over PCIe (similar to implementations of pseudo-Ethernet over virtualization boxes). This is the primary assertion that the portability of existing applications via API tools is a straightforward exercise for software developers. Aggregate throughput increases linearly with the number of processing boards – achievable under a standard socket IP API, without requesting any long integration effort from the user. The overhead burden taken by the API’s kernel threads remains extremely low, leaving 96 percent of the processing power available for the application. In turn, use of the IP socket API protects the end-user’s software investment for the long term – essential to extended military deployments. API tools can be as simple to use as standard switched Ethernet, while allowing higher performance and lower latency; also cost of ownership improves with Ethernet-like usability.
Executing a path to VPX
Routine tech refresh opportunities must not only capitalize on technology advances but also address DoD mandates for achieving the most effective tactical capabilities in the face of budget challenges. As a result, VME-to-VPX refresh strategies must take advantage of standards-based tools that speed deployment. The manufacturer’s objective is to enable a smooth shift from a GbE TCP/IP implementation, avoiding low-level complex and proprietary APIs (complex custom software interfaces) typical of many Serial RapidIO implementations. At the same, the migration should take advantage of major performance breakthroughs and higher bandwidths as compared with GbE.
As a result, OEMs benefit from an optimized Total Cost of Ownership (TCO) and have an extremely cost-efficient migration path from VME into the next big thing in VPX: 10 G and 40 G Ethernet communication on the backplane. Most importantly for VME platforms in stages of tech refresh, designers must consider performance levels of deployed systems – leveraging data flow tools such as APIs to simplify migration to VPX and establish a critical path forward toward increased bandwidth, performance, and flexibility.
David Pursley is a Product Line Manager with Kontron. He is responsible for Kontron’s VPX, VME, CompactPCI, MicroTCA, and AdvancedTCA product lines in North America and is based in Pittsburgh, PA. Previously, he held various positions as a Field Applications Engineer, Technical Marketing Engineer, and Marketing Manager. David holds a Bachelor of Science in Computer Science and Engineering from Bucknell University and a Master’s degree in Electrical and Computer Engineering from Carnegie Mellon University. Contact David at [email protected].
Kontron 888-294-4558 www.kontron.com