System Architectures for High-Channel-Count or High-Performance Applications

When implementing a high-channel-count or high-performance application, give careful thought to the system architecture. The system architecture will affect not only the current performance of a system, but also the ability to expand the system in the future.

This document discusses several basic types of system architectures that you can implement around PXI modules, and looks at the advantages and disadvantages of each.

PXI Embedded Controller and PXI Chassis

In this architecture, there is a single PXI chassis with a PXI controller. This is the simplest and most common architecture. The PXI controller is the limiting factor for things like processing speed and streaming to disk, while the size of the chassis determines how many modules can physically fit in the system.

Advantages:
· Simplest hardware and software implementation
· Small form factor

Disadvantages:
· Might need more slots than available in a single chassis

See Also:
What is PXI?

PC with MXI-4 to PXI Chassis

One way to achieve higher performance is to move to an architecture using an external PC. While PXI controllers may be high performing, they don't always offer all of the configuration flexibility that PCs do. Multiple video cards, dual processors, RAID configuration, etc. are all configurations not supported in PXI controllers. By using a MXI-4 link (PCI-PCI bridge), you can connect a PXI chassis to a PC. This enables you to take advantage of the high channel count instrumentation of PXI as well as all the benefits and flexibility of the desktop PC.

Advantages:
· PCs offer greater performance
· Greater variety of accessories for the PC, such as a RAID controller
· Lower cost than a PXI controller

Disadvantages:
· Requires more space than a PXI controller
· Not an industrial form factor

See Also:
PXI vs. Industrial Computers
PCI Express - An Overview of the PCI Express Standard

Multiple PXI Chassis with a PC or PXI Controller

Both the embedded controller and PC architectures are limited by the number of slots available in a single PXI chassis, but many applications require more than the 17 modules usable in the largest PXI chassis. One way to increase the channel count of the system is adding more PXI chassis to the system.

You can add additional chassis with MXI-4 in either a daisy-chain or star configuration. In the daisy-chain configuration, connect the PXI chassis in series with each other. Every time data crosses a PCI bridge, the expected total data throughput across that bridge is cut in half. This restriction in data throughput in the daisy-chain configuration is caused by unequal PCI bus arbitration between all of the slots and can be particularly important in high throughput applications. One way to get around this limitation is to use a star configuration. In the star configuration, the PXI chassis are all connected directly to the PC. This means that the controller has several MXI-4 links to it, each connected to a single chassis. In this case, the maximum throughput to each individual chassis is the same and limited only by the total throughput the controller can sustain. By using a PC with multiple independent buses, you can increase the aggregate throughput that the system can handle when in a star configuration.

Advantages:
· Can achieve higher channel counts

Disadvantages:
· Increased channel count means increased demands on the controller

Modular Architecture

The modular architecture breaks the system into nodes, where each node is a self-sufficient system in itself. A modular architecture is very scalable when it comes to channel count. You can add another node to the overall system in order to increase the channel count.

The modular architecture brings some added complexities. Each node can function independently; however in a large system, it is key that all the nodes function together to meet the overall system requirements. You must coordinate hardware and software synchronization through a designated “master” node. While all of the slave nodes might be exactly the same in both their hardware configuration and application software, the master will have any extra hardware needed as well as a different software implementation.

For systems with only a few nodes, the master might also double as a functional node. In larger systems, the master is a node dedicated solely to controlling and synchronizing the entire system.

Advantages:
· Highly flexible and scalable

Disadvantages:
· Most complex implementation of hardware and software

See Also:
Building a multi-client single server distributed architecture
Implementing Configuration Files for Computer Based Data Acquisition Systems

Reader Comments | Submit a comment »