RTC Magazine

Many industries that have traditionally relied on VME are driving to transform and modernize their systems to take advantage of new processor architectures and fabric-based system architectures as quickly as possible. This poses a challenge because system integrators and OEMs would like to make incremental changes to existing systems and not have to redesign all elements. They would like to reuse I/O that has traditionally been supported on VME, while updating the compute control core to new technologies and have them work together. Rather than selecting a single standard, the optimal solution will be provided through support by both VME and MicroTCA platforms for common, published, open software standards in terms of operating systems and middleware. It is this convergence of existing and new technologies that will be critical to the modernization of high-end systems.

MicroTCA leverages the emerging ecosystem of the Advanced Mezzanine Card (AdvancedMC) to create a new, flexible, small form-factor platform. It is important that system manufacturers and integrators understand the capabilities that MicroTCA brings to the embedded market.

AdvancedMCs were originally developed for AdvancedTCA (ATCA) platforms. ATCA has been targeted at the “aggregation” layer of the great “telecommunications network onion,” in between the outer edge “access” devices and inner “core” switching functions. Here, the priority is placed on a high compute capacity per blade, high availability, high bandwidth fabric between the compute blades and a few high-speed, high-density network I/O interfaces. And, given the investment in ATCA chassis and blades, it is important to have an ecosystem where many different solutions can be built around the same fabric (and thus the same or compatible blades).

AdvancedMCs bring further economies of scale to ATCA in the form of a more granular approach to the addition of either specialized processing elements or I/O interfaces. Adding from one to eight AdvancedMCs to an existing ATCA processing, I/O or dedicated carrier blade incrementally extends the solutions possible from a single ATCA chassis, both horizontally toward the network edge and vertically within the aggregation layer. This leverages the investment in ATCA into wider telco applications.

However, there are still environments where ATCA is simply either too large or too expensive to serve as a base architecture, or where -48 VDC is not an option. Thus, the idea of MicroTCA as an architecture was born (around the same time AdvancedMCs were first conceived), where the same pay-as-you-grow scalability could be realized in a much smaller, self-contained carrier assembly, and where, to a certain extent, the chassis form-factors (and applications) are limitless. Here, a typically passive backplane carrier can be mated with the appropriate type and number of fabric “hub” modules to provide the bandwidth, protocol and redundancy required for the application.

Six AdvancedMC form-factors of varying component height and module width have been specified, all leveraging the same high-speed, 170-pin edge connector. The MicroTCA specification allows for modular or monolithic chassis configurations from one carrier and one module to 16 carriers and 192 modules, while ensuring that modules always see the same “virtual” environment. These MicroTCA communications servers typically support two to three independent fabric interconnects on a carrier, where each fabric “port” (differential transmit/receive pair) is capable of up to 6.25 Gbits/s in each direction, and specific ports can be aggregated to form “fat pipes” with higher throughput.

PICMG has defined AdvancedMC fabric interconnect standards based on Gigabit Ethernet/10 Gigabit Ethernet (GigE/10GigE; AMC.2), PCI Express (AMC.1) and Serial RapidIO (AMC.4), along with defining storage interconnections based on SATA/SAS or Fibre Channel (AMC.3). The MicroTCA specification leverages these standards, allowing for switched and/or point-point fabrics. Current MicroTCA fabrics typically range from 1 Gbit/s (one port) to 12.5 Gbits/s (4 ports). The current aggregate carrier (switched backplane) bandwidth is around 40 Gbits/s, but next-generation hubs should exceed this; and the MicroTCA specification allows for up to 12.5 Gbits/s per port.

At present, AdvancedMC modules are available or under development for general-purpose processing (x86 & PPC), Digital Signal Processing (DSP), Digital Signaling (E1/T1/J1, OCx, DS3, etc), Serial ATA (SATA) and Serial Attached SCSI (SAS) storage, GigE/10GigE, Wireless Broadband (WiMAX), Voice-over-IP (VoIP), VGA video, and even PCI/PCI Telecom Mezzanine Card (PMC/PTMC) carriage. While the majority are focused specifically on telecommunications, more than a few provide applications in other solution spaces, such as military, aerospace, industrial and medical.

AdvancedMCs were designed to integrate into the highly available environment of ATCA, and this support carries over to the MicroTCA environment. Offering hot swap, Intelligent Platform Management Interface (IPMI), dynamic fabric negotiation, power budgeting and more, MicroTCA covers the availability, serviceability and manageability requirements of many target markets. In addition, unlike the ATCA blade fabric, AdvancedMC fabrics can differ between MicroTCA (and ATCA) carriers (within the same or different shelves), allowing the most appropriate choice for the application. As such, the lower overall cost of these modules will allow a less fabric-sensitive ecosystem to develop.

VME/MicroTCA Convergence

At first blush, MicroTCA appears to compete with VME, especially with next-generation VME+ fabric solutions, such as VXS (VITA 41) and VPX (VITA 46). While there is clearly a choice that must be made when considering general-purpose computing platforms, VME continues to be the logical choice in many military, industrial control and medical imaging applications. Because VME has a large ecosystem of COTS and custom I/O targeting military applications, it will continue to be a critical architecture for many years to come.

The purpose-built backward compatibility of each successive revision to the VMEbus standard allows many VME edge systems to be refreshed with single board computers that support POSIX operating systems, open middleware, 2eSST and multiple GigE interfaces. These SBCs can integrate into environments with ease, yet still continue to interface with legacy I/O devices. In many cases, they can also communicate internally at fabric speeds via 2eSST—all in an existing chassis.

However, some of the traditional roles VME has played—such as concentrated digital signal processing systems and compute centers—can migrate to alternate technologies such as network-centric MicroTCA communications servers. A heterogeneous blend of VME-based I/O subsystems supporting traditional VMEbus I/O will exist, primarily at the edge of the system, interfacing to weapons or sensors. Concurrently, compute-centric functions, along with some digital signal processing systems, can take advantage of the compute density of MicroTCA.

How Will it Work?

Software standardization efforts will allow the development of heterogeneous environments where common software architecture is used throughout the system. VME and MicroTCA-based systems will run common operating systems and middleware optimized for that architecture. Applications running on VME processor blades will share data through the middleware, using plug-in modules that optimize the transport for 2eSST, while MicroTCA systems will have transport plug-ins optimized for GigE or PCI Express. Data sharing between systems will continue to use GigE to share data between applications running on different systems. Abstracting to a middleware product, like DDS, eliminates the need for application software to be written for one particular system architecture. Contractors can evaluate new technologies, such as MicroTCA, as candidates for upgrading hardware, while architects designing new systems can pick and choose the most appropriate architecture for each part of the system (Figure 1).

New technologies such as WiMAX (IEEE 802.16) can also be deployed using MicroTCA communications servers. WiMAX is a wireless broadband technology that is being considered for systems that will help control the cost of wireless high-speed transport. Platform variants defined in the MicroTCA standard, like cubes and Pico shelves, are well suited to support network nodes based on WiMAX. At the same time, rugged variants of MicroTCA communications servers will move MicroTCA into the field and into vehicles.

In order to take this one step further and to determine how AdvancedMC blades and MicroTCA chassis components can be ruggedized for a variety of environments, a Special Interest Group (SIG) has been formed by a collection of companies interested in MicroTCA. The goal of the SIG is to allow the use of COTS AdvancedMC modules in both commercial and military/industrial applications. In addition, the “ruggedization” SIG is researching how new packaging techniques can be applied to commercial AdvancedMCs to harden them sufficiently for harsh environments and whether conduction-cooled systems can be designed around them. If COTS AdvancedMCs can be used, the overall system cost will be lower and the variety of blades available to an application will dramatically increase. The work of the SIG will be part of an official standards body working group, so the completion of a ruggedization specification should coincide with the release of the MicroTCA standard.

VME64x still offers a great deal of variety, but at reduced backplane bandwidth. VITA 41 will increase the backplane bandwidth with the addition of fabric interfaces, but this will require a system retrofit. MicroTCA has a much smaller footprint and the ability to scale better than a VXS-based system. In addition, the AdvancedMC ecosystem has already begun to grow prior to the release of the MicroTCA system standard.

The use of AdvancedMCs in a variety of applications both on ATCA carrier blades and within MicroTCA communications servers will help to substantially drive down their cost. Moving forward, future systems will be able to leverage AdvancedMC flexibility, without the added expense and infrastructure requirements of ATCA. Furthermore, applications that can benefit from the increased power envelope of ATCA can also benefit from the variety of AdvancedMCs that will be available for both environments.

Motorola Tempe, AZ. (800) 759-1107. [www.motorola.com/computing].