Many communication systems have the same requirements as military data processing and digital signal processing systems. Often commercial solutions can be used to solve challenging packet processing and digital signal processing (DSP) problems in military applications. Many sensors used for signal processing collect and process the signals before packetizing the data, similar to the way signals are processed and packetized in communication applications.
In the past, these processing functions have often required special-purpose hardware such as discrete Network Processing Units (NPUs), digital signal processors (DSPs), co-processors, or field-programmable gate arrays (FPGAs). However, recent enhancements to Intel® architecture processors, together with advanced software, are providing developers a viable alternative, whereby they can use a single blade architecture for consolidation of all their application, control, signal, and packet processing workloads on Intel® architecture. Intel processors such at the Intel® Xeon® E5-2400/2600 are ideal choices for data intensive packet processing.
Improvements in multi-core architectures combined with the latest DSP capability in Intel’s Advanced Vector Extensions (Intel® AVX) and the packet processing software enhancements provided by the Intel® Data Plane Development Kit (Intel® DPDK) make Intel® architecture increasingly attractive. Additionally, by consolidating DSP and packet processing with other workloads on a Intel® multi-core processor, it is possible to reduce hardware costs, simplify the application development environment, and reduce time to market—with all the factors combining to reduce overall total cost of ownership.
Emerson Networking Power Embedded Computing, a Premier member of the Intel® Intelligent Systems Alliance, with extensive experience in the telecommunications industry and the defense industry, has seen the potential for using Intel processors combined with AVX and the DPDK in the role of a signal and data processing. At Embedded Tech Trends 2013, Emerson Networking Power Embedded Computing proposed a model using an AdvancedTCA blade-based platform as a solution for challenging DSP and data processing military applications.
They list the following benefits to using AdvancedTCA for DSP and packet processing:
- AdvancedTCA blades efficiently supply many processing cores and adequate memory.
- Multi-core Intel® Xeon® E5-2400/2600 processors that are well suited to process complex data.
- 40G Ethernet direct connection between the AdvancedTCA fabric and the processors
- Inherently rugged nature of blade-in-chassis design suitable for rugged benign environments such as shipboard, manned airborne, or transit case applications.
- AdvancedTCA is an open standard with many suppliers.
The Emerson Networking Power Embedded Computing model combines the best of AdvancedTCA with the Advanced Vector Extensions for performing DSP algorithms and the DPDK for data packet processing to construct a platform ideal for demanding signal processing.
Inbound packetized sensor data enters into the AdvancedTCA switch over 10G (40G in the near future) fabric where flow control software on the ATCA switch flows the data to the appropriate processor board that can perform the DSP algorithms before being routed over the fabric to outbound destinations.
In a typical scenario, an application running in the control plane of the processor board, along with flow control software ensures that the data is directed to the correct destination. Algorithms to process the data are assigned and dispatched to specific threads on specific cores within the processor by the control plane application. Packet processing extensions defined by the DPDK eliminate interruptions in the operation of the algorithms. With the assistance of AVX, the DSP algorithms are completed without interruption. The control plane application steps in again to direct the results to the next destination. For instance, to another thread running a different algorithm, to a different processor board, or out of the system through the fabric switch.
The main goal of the Intel® DPDK is to provide a fast simple framework for fast packet processing in data plane applications. The DPDK is freely available from Intel to developers. Developers may use the code to understand some of the techniques employed, to build upon for prototyping or to add their own protocol stacks. Alternative ecosystem options that use the Intel® DPDK are available.
The Intel® DPDK is designed to operate with Intel® Xeon® class processors. It provides optimized data plane libraries and optimized drivers that operate in user space. It operates under Linux, which separates high level control functions from algorithms running as threads on specific dedicated processor cores, splitting the load to run efficiently in the multi-core environment. The DPDK has a low overhead run-to-completion model that is optimized for the fastest possible algorithm performance.
Additional libraries and drivers in the Intel® DPDK provide support for data packet processing:
- Memory Manager (huge page tables to optimize performance)
- Buffer Manager (optimized memory allocation tool that eliminates need to lock)
- Queue Manager (manage incoming and outgoing data to the cores)
- Flow Classification (IP flow management, optimized around Ethernet controller)
- Poll Mode Drivers (user mode drivers eliminating interrupts for threads running algorithms)
While the DPDK handles data packet issues, Intel ’s Advanced Vector Extension (AVX) addresses the digital signal processing demands. AVX extends 128 bit SIMD instructions to 256 bits. This potentially doubles floating-point operation performance when using single precision floating-point numbers. Each processor core supports AVX instructions so that the DSP algorithms can be assigned as necessary.
Intel supplies optimized libraries, Integrated Performance Primitives (IPP), for AVX, while optimized Vector Signal Image Processing Libraries (VSIPLs) are available from Intel® Intelligent Systems Alliance members.
The Wind River Systems portfolio of network acceleration solutions is optimized for Intel’s packet processing architecture. Wind River Linux* has built-in high performance optimizations for the Intel® Data Plane Development Kit (Intel® DPDK).
6WIND SA provides packet processing software to meet both the wire-speed performance and time-to-market requirements of mobile infrastructure, network security, high-frequency trading, and deep packet inspection applications. 6WINDGate* typically delivers up to ten times the performance of standard OS networking stacks while maintaining full compatibility with standard application APIs. 6WIND provides the Intel® DPDK libraries pre-integrated within the 6WINDGate software, further accelerating the development process for OEMs.
NASoftware Ltd, in collaboration with Intel, has development conversion tools to help programmers switching from PowerPC/AltiVec to Intel® Streaming SIMD Extensions (Intel® SSE) processors and Intel® Advanced Vector Extensions (Intel® AVX) processors.
LEARN MORE >>
- Roving Reporter: Accelerating time-to-market for OEMs using Intel® DPDK library
- Intel® Data Plane Development Kit: Getting Started Guide
- Packet Processing on Intel® Architecture
- Signal Processing on Intel® Architecture
- Signal Processing with Intel® Advanced Vector Extensions