Consolidating Railway Communication
Virtualization can break down the barriers between various non-vital rail travel applications to cut costs and effort.
The number and variety of different automated services in modern rail travel has proliferated into a cost and management nightmare. Using Intel® Virtualization technology to bring these applications together into a unified environment can greatly increase efficiency and reliability, while reducing costs.
Modern railroad networks are a prime example of complex, rugged environments that require multiple computer systems for management, control, and convenience. These requirements go beyond the vital systems charged with controlling engines, fuel, braking, and more. Today’s rail transportation also includes IT-type applications for energy management, passenger information systems, door control, ticketing, video surveillance, entertainment, and on-board Internet, to name a few.
In addition, all these applications must communicate and exchange data at certain levels. Until now, this set of applications had been handled by separate and distinct computer systems. But the increased complexity of harnessing these isolated systems is growing more expensive and less manageable by the day.
Not only does each independent application have its own hardware, but most likely also requires a unique operating system and application software. This means railway management must deal with a host of different suppliers as well as communication and interoperability issues between systems. (Figure 1).
And it means difficulties with certification and obsolescence for each system, with some systems having far more capacity than they actually use, which translates to higher costs. For example, a ticketing system operates sporadically, and mostly while the train is in the station, so its dedicated computer sits idle the majority of the time.
If it were possible to bring these many disparate applications together into one computing environment—where they could share resources, exchange data more easily, and still fulfill their specialized functions—the savings in time and cost could be truly significant. To do this would require virtual computing environments, where each application could be transported with little or no alteration, and still run as if it were in its familiar old environment. Virtualization technology enables this scenario by networking virtual machines running different operating system environments and applications on a single, common hardware platform.
One such system family utilizing virtualization for just this purpose for modern railway computing is the menRDC. It combines different functions needed for an IT infrastructure on railway systems in a single, configurable and rugged package, providing a main server, storage system, and a network Ethernet switch. At the heart of the menRDC main server is a ruggedized CompactPCI Serial SBC with an Intel® Xeon® D-1500 (Broadwell) system-on-chip (SoC). This multicore hardware platform allows for the running of multiple virtual machines supporting different operating system environments. In addition, the SBC provides a rich combination of I/O interfaces including PCI Express, SATA/SAS, USB, and Ethernet plus signals needed for general system management. (Figure 2)
The modular virtualization-based system is available in several pre-configured modules but can also be configured as a “built-to-order” system, thanks to the compatibility of the computing, networking, storage, and communication modules available.
For example, the main server, MEN Micro’s MH70R, consists of one or two CPUs, each with a 16-core Intel Xeon D-1500, 2×10 Gb and 4×1 Gb Ethernet 3G/4G, Wi-Fi and GPS modem slots, up to four hot-swappable SATA slots and a pre-configured Linux operating system with drivers. Along with network switching and expanded storage units, these components can be combined in single- or multiple-rack systems with standard or custom configurations to meet a wide range of processing and capacity needs (Figure 3).
Virtualization Brings It Together
Given the power built into such a hardware design, how do we bring an array of applications together into this single platform, when many of the applications are running on different processors and operating systems? The answer is virtualization, which allows us to set up different execution environments, called virtual machines or VMs, on a single platform, sharing the underlying physical resources of that platform. The trick is to be able to do it efficiently, while maintaining the level of performance demanded by the various applications. The railway system described above does this with the combination of a lean and efficient type 2 hypervisor along with Intel’s hardware assisted virtualization technology (Intel® VT).
The hypervisor abstracts the hardware from the application. It is hosted by Linux as a regular application and relies on the services of that operating system to manage system resources (CPU cores, memory, I/O, storage, etc.) for the applications it hosts in the various VMs it provides. Each VM has its own private virtualized hardware used by its application, as if it were on a separate processor environment. In reality, it’s running on one or more of the D-1500’s cores through the hypervisor. In addition to running the application, the hypervisor can be adapted to monitor other tasks, such as load balancing and crash protection. (Figure 4)
Any such virtualization involving a hypervisor is going to involve overhead, which will affect performance. Meeting performance demands of the applications is accomplished in two ways: First is the raw processing power of the D-1500 family, with eight cores on the D-1539 and 16 on the D-1577. The more cores, the greater the potential for workload consolidation. The other is Intel’s hardware virtualization technology, which simplifies the software—and hence the overhead—used to enable and manage virtual machines.
The silicon virtualization technology addresses three areas. VT-x focuses on execution cores and CPU virtualization to reduce hypervisor complexity. VT-d focuses on direct memory access, remapping DMA transfers and interrupts for efficiency where the guest application is unaware of the physical addresses. VT-c focuses on Ethernet connectivity, such that network devices are aware of VMs and will have Rx/Tx queues dedicated for each VM, which reduces a great deal of remapping without involving the hypervisor.
Optimized Railway Networks
The combination of rugged, modular and powerful multicore hardware with efficient virtualization technology in both software and hardware results in a single platform running many applications. Applications that formerly ran on different, dedicated hardware platforms are freed from hardware dependence and can share resources while running, as happens in the menRDC platform.
The savings to maintenance and upgradability are significant. If greater capacity is needed, the system can easily be increased by simply installing additional modules. The applications can efficiently share information as needed. Management can devise a user interface that conveniently accesses all needed data and control functions. Thanks to virtualization technology, modern railway systems can work on a common platform, reducing integration complexity and obsolescence issues, while better streamlining the needed system performance.
Gunther Gräbner has working in the automotive industry for more than eight years as a hardware developer, technical project leader and team leader of hardware development. Since 2011, he has supported MEN Mikro Elektronik GmbH as product manager for customer projects and MEN standard products.