"Multi Node Architecture" Offers an Innovative Alternative
Combining PCI, cPCI, and custom concepts into a cost-effective, non-proprietary chassis system, Tracewell's Multi Node packaging technique achieves many benefits.By Melissa Lander, Tracewell Systems, Inc.
Design engineers who specialize in telecom and datacom applications are under increased pressure to implement designs satisfying a broad mix of critical design factors. These include such criteria as performance, physical size, power consumption, ruggedness, reliability and serviceability, and of course, cost. Increased performance requirements and shorter time-to-market have resulted in fierce competitive pressures that are moving the industry away from custom and proprietary designs. The solution to this problem begins with the most fundamental decisions the designer makes, including the choice of a computing platform.
Many telecom/datacom systems are based on PCI (Peripheral Component Interconnect) architecture in standard PCI, cPCI (Compact PCI), or PMC (PCI Mezzanine Card) form factors. All these solutions adhere to the same electrical standard, although they are mechanically non-interchangeable. PMC cards plug into card carriers within the system. Designers who opt for Compact PCI components can rest in the knowledge they have selected a robust architecture developed for industrial system requirements. However, the designer must also justify the relatively high cost of cPCI, which exceeds that of standard PCI, but is still less than custom hardware. Where cost does become the governing factor, the design engineer may be forced to compromise on hardware, and turn to standard PCI components.
The traditional design procedure with off-the-shelf PCI components has been to select cards according to I/O, processing, and other requirements, and then fit them to an enclosure. However, such components, and thus, the final system, tend to be generic to a great extent. Therefore, the system designer may find it challenging to incorporate some of the more specialized features of cPCI, such as the abundant power and cooling needed for power-hungry telecom peripherals.
Tracewell Systems specializes in electronics integration, and has historically taken the approach of packaging hardware in ways that also achieve maximum space efficiency, cooling, reliability, accessibility, or other critical requirements not easily realized with generic solutions. Recent telecom/datacom system trends clearly showed the need for a new chassis design that might combine the cost benefits of existing PCI technology with the robustness and connectivity of cPCI. Already experienced in the design and production of electronic systems packaging for VME, VXI, PXI, PCI, Compact PCI and proprietary bus systems, the company saw opportunities for improved cooling, EMI/RFI shielding, remote monitoring, and structural integrity that would add value to telecom/datacom system enclosures, and provide a fully-integrated, "Level 5" solution ready to plug in and power up.
Designing the Ideal System
The goals and requirements for an ideal telecom/datacom platform are actually independent of whether the design engineer chooses PCI or cPCI. Typically, these include:
Space Efficiency and I/O Density. Higher I/O density enables more bandwidth to be served from a given computing system or rack. This ultimately translates into more subscribers being served by a given installation.
Fault Resilience/Tolerance. In a business where revenue is calculated based on minutes of system up-time, the failure of telecom/datacom computing systems residing in difficult-to-access locations can seriously impact profits. Ideally, any such system should be able to instantly and automatically compensate for failures by relying on redundant capability. Most carrier-class equipment is expected to operate at a 99.999% ("five nines") reliability level, which equates to spending a mere five minutes per year off-line. A completely fault tolerant system can continue operating, even in the event of a CPU failure.
Maintainability and Expandability. Ongoing, rapid availability of spare parts and support is critical, but moreover, easy serviceability is equally important in assuring rapid repair. Lost revenue caused by low-cost solutions that offer poor factory support can wipe out any initial purchase economies in just a few minutes of down-time. System designs that provide for rapid access and easy disconnect of I/O terminations enhance serviceability. Systems should also be reconfigurable and scalable to permit optimization of capabilities as mission requirements change or grow.
"Hot-swap" capability. The ability to remove a card or subassembly from a system without powering down the entire system is extremely desirable because it facilitates maintenance without disrupting operation of those portions of the system that are still operating properly. Hot swap is a feature that can be used in cPCI.
Remote Monitoring and Management. Hardware in difficult-to-reach locations can be costly to visit for routine maintenance, as well as for repair. In such cases, the ability to remotely monitor site health can simultaneously enhance system reliability and save on maintenance expenses. By tracking real-time data from remote sites, service personnel can often detect trends in system performance that enable them to anticipate failures and practice pre-emptive maintenance. The addition of system control can enable technicians to reconfigure hardware remotely. Power and Heat Management. These design criteria have assumed a mission-critical role in assuring uninterrupted operation of equipment, and are clearly interrelated. With a best efficiency of only about 80%, power supplies contribute a significant portion of a system's total thermal load. Inadequate cooling of cards and power supplies leads to frequent equipment failures and shortened system life. User Perception. As stated earlier, end users may be sensitive to form factor or other cosmetic issues that have nothing to do with performance. To achieve a "telecom image," they may opt for a cPCI-based system at much higher cost, even though standard PCI might be adequate from a functional standpoint.
"Multi Node" Architecture
Even though cPCI architecture is highly regarded for its robustness and premium feature set, much of the same basic functionality nonetheless exists in standard PCI components. Tracewell studied the feasibility of combining PCI, cPCI, and custom concepts into a cost-effective, non-proprietary chassis system, and introduced the "Multi Node" concept in 2001.
Figure 1. Multi Node is a PCI packaging technique that achieves many benefits of Compact PCI, including modularity and hot swap capability.
Multi Node system architecture (Figure 1) consists of a 10U, 19 inches-wide, EIA-compliant rack mount chassis which is subdivided into a computer module section and a power/cooling section. The computer module section accepts up to twelve, 12 inches-wide, sub-1U computer modules. Each module is a separate, removable sub-chassis that contains card slots to accept two PCI cards and two PMC modules, in addition to custom circuitry (Figure 2).
Figure 2. Multi Node fits all the components needed to configure a complete telecom PC into a 12 inches-wide, sub-1U slide-out chassis.
The rear of each module is fitted with a power connector, and I/O connectors that satisfy application requirements.
The power/cooling section consumes the remaining system space, and houses power supplies and forced air cooling that service all computer modules. Systems are built in 10U segments, allowing further cost savings through standardization of structural components Multi Node uses common parts and modules, thus reducing stocking requirements. The system also achieves a telecom "look and feel" without incurring the additional cost of cPCI.
High Density, Remote Monitoring, and Custom Capability
A Multi Node computer module supports integration of custom features and high I/O density using cost-effective PCI cards. The height of a Multi Node computer sub-chassis is actually less than 1U. Up to twelve complete computer modules can be housed in a 10U rack, providing up to 20% more computing capability as compared to traditional 1U or larger form factors and other packaging strategies. All computer modules are supported by a centralized N+1 modular power and cooling section, helping to further reduce overall system cost while enhancing reliability and serviceability.
User-specified options can be added to a Multi Node computer module, such as Tracewell Systems' microprocessor-based Chassis Monitoring and Management (CMM) module. CMM integrates remote control, remote monitoring, data logging, power cycling, and event recording into the system, which can be indispensable where system installations are not easily accessible. While monitoring capability adds to system cost, the investment is repaid quickly. It enables service personnel to remotely monitor rea l-time data, and to tailor maintenance schedules to actual system health for improved maintenance efficiency and reduced down time.
Improved Reliability through Redundant Power and Cooling
A Multi Node system is powered by an array of redundant industrial power supplies in an N+1 configuration. Power supplies can be fitted to enable operation from DC or AC mains. In N+1 operation, the load is shared among multiple, identical power supplies. One supply can fail, and the rest can still carry the entire load. For example, a 1500 watt N+1 system could consist of three 750 watt power supplies, any two of which can fully power the system. High-power N+1 is a feature not easily implemented with a standard PCI chassis, which does not support high power/density configurations. However, it is common in industrial rack systems, and it is a standard feature of Multi Node. Redundant system cooling is supplied by the use of multiple fans in the power/cooling section of the Multi Node chassis (fan rotation can also be monitored by CMM). Segregation of power supplies from computer modules enables power supply exhaust heat to be kept clear of computer modules, simplifying cooling requirements in those modules.
Figure 3. A CMM Module provides comprehensive monitoring and management of an entire multi-slot chassis.
Other Advantages of Multi Node
In addition to the main advantages of versatility and enhanced reliability, Multi Node offers several secondary features and benefits, any one of which can be critical to an application.
First, Multi Node provides an easy-to-install, plug-and-go package that removes the burden of packaging design and integration from the end user. One concern with systems built piece-meal with conventional PCI chassis, power supplies, and cooling is that ultimate reliability may be known only after the system is placed in service. Multi Node carries with it 25 years of expertise in systems packaging for military/aerospace, medical, and industrial control.
Second, users have tremendous freedom to choose the exact functionality they need. The same Multi Node concept can be cost-effectively adapted to non-telecom applications, and the computer module can be modified to meet special user requirements. Wireless applications typically employ hundreds to thousands of individual systems, so the elimination or optimization of just a few components or features per module can result in substantial savings.
Finally, the Multi Node concept can be configured in a full 19 inches chassis to provide yet another lower-cost alternative to cPCI. This variant of Multi Node contains all boards, a power supply, and a cooling system in a space-efficient 1U chassis, for mounting in an EIA rack. This packaging concept provides many of the benefits of the split-chassis Multi Node design, including high density, remote monitoring/control capability, cost effectiveness, and a designed-for-telecom appearance.
Figure 4. The 1U version of Multi Node is a self-contained telecom package housing computer, power, and cooling in a 19" rack-mount chassis.
The Multi Node concept enables telecom/datacom system designers to realize many of the benefits of Compact PCI through innovative packaging of less expensive PCI cards. Most industry predictions indicate that wireless devices, now among the most prolific of all consumer electronics, will only grow in popularity. As the build-out of wireless infrastructure continues, and as 3rd Generation (3G) enhancements gain momentum, providers who find ways of increasing operating efficiency and system performance will maintain the greatest competitive edge.
Several specific benefits of the split-chassis and 1U Multi Node designs have been discussed in the preceding paragraphs, but the "bottom line" for Multi Node can be summarized as enhanced I/O density, reliability, serviceability, growth potential, and cost effectiveness. Functionally sophisticated systems can be built with such enclosures at a lower initial cost per node. Savings can be further compounded over time in the form of more efficient maintenance and an easier migration/upgrade path as requirements change and technology progresses.
Melissa Lander is a Marketing Coordinator for Tracewell Systems, Inc., 567 Enterprise Drive, Westerville, Ohio 43081. She can be reached by email at firstname.lastname@example.org. For more information on the company, visit www.tracewellsystems.com.