Radar's brave new world: OSA
High-Performance Embedded Computing (HPEC) is paramount within the tight real estate confines of defense systems today. HPEC’s high-scale parallel processing technology, based on Open Systems Architectures (OSAs), also increases the power of sensor applications such as radar.
But high performance has not always gone hand in hand with openness. Until recently, the best results in radar were achieved through proprietary software and hardware, which customers preferred even though it might lock them into particular vendors. Programming was difficult because the parallel execution so necessary for radar processing had to be built in at the application level. If six processors were required, six programs had to be written, each of which made calls to the math and communication libraries. These programs were difficult, time consuming, and expensive to write as well as challenging to upgrade. But these days OSA- and open standards-based software tools can change this paradigm.
Radar: Shift to OSA
As technology evolved – spurred by the consumer market – Commercial Off-the-Shelf (COTS) and OSA hardware and middleware products (as well as the toolkits supporting them) reached the volume and acceptance level required to support shifts in the design of compute-intensive applications such as radar. Open architecture means that necessary programming interfaces are available and published, giving rise to a greater wealth of competing solutions, easier interoperability, and lower life-cycle costs as compared to proprietary solutions. High-performance middleware also means that some of the parallelization required by sensor applications – and previously done manually – can now be executed automatically through such aids such as signal processing and math libraries.
At the programming level, OSA approaches have caught up with proprietary systems through middleware such as VSIPL++, the latest version of the Vector Signal Image Processing Library. Application development toolkits have incorporated such middleware and other advances to make it easier for customers to use these OSA technologies.
The Pentagon acknowledged these trends in 2010, when Under Secretary of Defense Ashton Carter called for the use of OSA as part of acquisition reform. This high-level support has flowed down to new procurements and upgrades. After all, since new hardware will make proprietary applications obsolete, customers might as well migrate to open systems middleware, too. Additionally, radar systems already had begun to reflect the technology shift before the Carter memorandum. MIT Lincoln Lab, for example, has employed the Radar Open System Architecture (ROSA), involving COTS hardware and open systems. ROSA has been applied in the prototype Cobra Gemini radar and the modernization of radars at the Kwajalein Missile Range.
The emphasis on OSAs in radar development continues. Requests For Information (RFIs) for new radar systems and upgrades are asking for open standards middleware technologies such as VSIPL++ and the open source Linux operating system.
Software development toolkits with OSA
One obvious way to lessen the difficulties of moving from a proprietary to an OSA-based radar system is to use software development suites with built-in OSA standards. These frameworks help developers meet deadlines by using math and signal processing functions such as the Fastest Fourier Transform in the West FFTW), VSIPL and VSIPL++, and data movement libraries such as Message Passing Interface (MPI) and Data Distribution Service (DDS).
Toolkits for radar system migration or design are available from multiple vendors. GE’s Advanced Multiprocessor Integrated Software (AXIS) environment, for example, includes math libraries, visualization tools, and interprocessor communications optimizers that run on multiprocessor boards – even hybrids combining traditional and highly parallel chip architectures (Figure 1). The software suite also gives programmers access to standard middleware such as MPI, OpenFabrics Enterprise Distribution – or OFED, and VSIPL++. (MPI integration into the AXIS toolkit is expected in the near term.)
Toolkits also can accelerate radar application development by helping to partition code across potentially hundreds of processors. Then programmers can assess software performance visually, viewing a system as it is running to check the efficiency of the code and whether it’s meeting its real-time constraints.
Low pain, high gain
The transition from proprietary systems to open systems architectures, assisted by the latest standards-based software tools, should be relatively painless for radar systems engineers and should lead to greater system performance, programmer productivity, and cost-effectiveness.