ICP Electronics
ITouch Partner
Lucent Partner
Marconi Partner
Nortel Partner
RAD Partner
Telco Partner
Tellabs Partner
VoxTechnologies 3Com Partner
Adtran Partner
Alcatel Partner
CA Partner
ChannelBank
Cisco Systems Partner
Eastern Partner
Enterasys Partner
Extreme Partner
Larscom Partner Kentrox Partner

At the core of this paradigm shift are two key concepts. The first concept is that manufacturing companies need pervasive networking. Virtually every person, machine, control, and sensor associated with bringing products to customers must be networked together to achieve the quality, time-to-market, and operational efficiency demanded of today's, and particularly tomorrow's manufacturing companies.

The second concept is that manufacturing companies need application-aware networking. Perhaps more than any other industry, manufacturing companies must transmit information from one place to another, where the fundamental form of the information differs tremendously from one application to another. Widespread use of computer models, simulations, client-server enterprise management, videoconferencing, document management, and email are but some of the critical applications with completely different networking requirements.

Networking vendors have advanced their product offerings over the past few years in part to address the requirements of manufacturing industries. Solutions have evolved from bridges and routers, to layer 2 switches and routers, to the new generation switch routers (also referred to as layer 3 switches). Although the product offerings from many vendors provide significant advances in satisfying various industries' networking requirements, the key to providing solutions for next generation manufacturing companies is completely addressing the two concepts previously outlined. This amounts to making "application-aware switch routing" available throughout a manufacturing organization.

At the other end of the model is layer 1, the physical layer. An example of this is the twisted pair cabling and associated driver and receiver hardware/software necessary to implement Ethernet. Layer 1 is associated with connecting one endstation to another.

Until very recently, networking equipment dealt only with layers 1, 2, and 3. Bridges, for example, provided layer 1 and 2 support, with an emphasis being on the intelligence associated with layer 2 (e.g. the MAC addresses of Ethernet). Routers, on the other hand, also supported layer 3 intelligence via implementing protocols such as IP and IPX.

The support of layer 4 in a networking device is extremely new, and it is a powerful feature as described later. It is in fact the end-to-end prioritization of network traffic based on layer 4 information which is the focus of this paper.

As networks grew, however, it quickly became obvious that an advanced form of network segmentation was required. This resulted in routers. Routers provided a way of dividing a network into groups, based on the concept that most traffic would be within a group (i.e. a bridge). Whenever traffic needed to go outside of a group, it would go through the router, which used layer 3 technology (e.g. an IP address) to determine where to send the information.

Bridges were fast, inexpensive, and simple, but the need for additional bandwidth eventually triggered the introduction of layer 2 switches which essentially performed the functions of bridges, but with far more intelligence.

Because of this significantly increased intelligence in switches (compared to bridges), there is less of a need for network segmentation. In fact, some very large scale networks have been implemented without any routers, and these are referred to as "flat" networks. However, in many situations, routers are still needed to connect switches together, particularly if wide-area-networking (WAN) is involved. Since WAN connections, using technology such as leased lines or frame relay, typically provide very small bandwidth, it is critical to minimize the traffic sent across such connections. Thus, even now with layer 2 switches being quite prevalent, routers are still used in many situations.

Like bridges, switches tend to be fast, inexpensive, and simple. However, routers are slow, expensive, and complex. And the reality is they haven't evolved significantly since they were introduced many years ago. They are perceived as the bottleneck in many networks in use today. This has triggered the introduction of switch routers.

As the "switch router" name suggests, switch routers essentially combine the functions of a (layer 2) switch and a router in one box. Combinations of ports can be treated as switched (i.e. layer 2 technology is used to forward packets between ports). Similarly, routing can be enabled between ports to gain the advantages of layer 3 network segmentation where appropriate.

Switch routers offer major advantages in many situations. Like switches, they typically utilize hardware-based packet forwarding engines. This means they are very fast compared to software-based routers, even though they may be implementing algorithms that are actually routing algorithms. Switch routers also tend to be far less expensive than software-based routers. This again is largely due to the fact that they are based on specialized hardware (Application Specific Integrated Circuits, or ASICs) rather than a complex general purpose CPU/memory/software architecture.

In theory, switch routers can replace both switches and routers. However, due to a difference in price-per-port, switches will still be used in price-sensitive situations where a layer 2 switch provides sufficient capabilities. Switch routers, however, will dramatically replace routers since there are very few downsides to switch routers over routers (support for legacy protocols being one advantage for routers).

Because switch routers are quite new to the industry, the ports supported tend to be primarily those that are in widespread use today. This typically includes at least 10Mbps Ethernet, 100Mbps Fast Ethernet, and 1000Mbps Gigabit Ethernet. Depending on the switch router, ATM and/or WAN ports may also be supported to allow the switch router to address large geographies.

In terms of layer 2 functionality, most switch routers support standard transparent bridging, which goes back to the original network bridges. Most switch routers also support the emerging 802.1Q standard for multi-vendor layer 2 support of network segmentation via VLANs. Operationally, switch routers generally have a high speed backplane of some type which they use to forward traffic from one port to another. Switch routers that are able to forward such traffic as fast as the ports can operate (e.g. at Gigabit Ethernet speeds) are referred to as "wire speed" switch routers. This forwarding speed can be one of the differences between one vendor's switch routers compared to another. If the speed of the switch router is critical (e.g. if it is at the core of a network), then this characteristic may deserve consideration when selecting one vendor's product over another's.

More so than the switching performance, however, are potential differences in the layer 3 route processing. Switch routers offer the ability to perform standards-based routing in order to forward packets. This typically includes support for the popular IP, and may also include support for Novell's IPX which is still very widely used in most manufacturing environments. Two key differences from one vendor to another are a) the degree of decentralization of the route processing, and b) the performance of the route processing with advanced features turned on.

The degree of decentralization can be key to a manufacturing company since this essentially dictates whether or not there is a single point of failure which can bring an entire network down. In most cases, manufacturing companies should look for switch routers with architectures that provide highly decentralized route processing with redundancy and hot-swappability.

One of the most distinguishing general characteristics of switch routers, however, are their performance when advanced features are enabled (and of course exactly what those advanced features are). Recall that the job of a switch router is to replace a router. As such, features found in modern routers such as Access Control Lists (ACLs) to perform filtering and security functions are typically important to manufacturing companies. Ideally, switch routers should have such features. And ideally, when those features are enabled, performance should remain "wire speed". In reality, many switch routers are lacking in such features, or at best they offer them but with substantial performance degradation when enabled (similar to what occurs in software-based routers).

A second form of layer 3 prioritization some switch routers support is referred to as IP flow (or layer 3 flow if protocols other than IP are supported). In some situations, this can in fact provide a fairly decent match to application-based prioritization. However, in others, it cannot. Prioritizing traffic by IP flows means that a given pair of IP addresses (source and destination) are given a certain priority. Thus, as an example, the flow between a user's PC and an email server could be given assigned one priority, while the flow between a user's PC and a streaming video server could be assigned a different (and higher) priority. Note, however, this mechanism does not work if there are multiple applications running on a single server. Having multiple applications per server is in fact quite common in manufacturing companies, and may be particularly true in smaller manufacturing businesses, or in branch offices of larger businesses. In fact, it may be extremely cost effective to combine applications on a single server, with the indirect downside of precluding the use of IP flows for traffic prioritization.

The unfortunate reality is there is no robust standard layer 3 prioritization mechanism currently available for multi-vendor environments. To truly achieve application-aware prioritization, networking equipment must resort to adding higher level intelligence, such as by utilizing layer 4 information.

OSI layers 1-3 are somewhat unique in that they are deliberately hidden from any knowledge of the application. This is by design so as to make these layers able to perform the same functions regardless of "what was going on above them". Layer 4 is the first layer where knowledge of the application exists. Using IP as an example, layer 4 is based on a "transport port" (often referred to as a "socket") which is generally assigned by application. There are many layer 4 protocols used in IP, but two very common ones are TCP or UDP, depending on whether a connection-oriented capability is needed (TCP is connection-based, UDP is connectionless). [Note that SPX is the typical layer 4 protocol associated with IPX.] IP layer 4 port definitions are specified in RFC1700. Examples are 20 for file transfer (FTP-data), 25 for e-mail (Simple Mail Transfer), and 80 for web browsing (www-http).

Note that if a switch router has the ability to look at layer 4 information, it can then perform intelligent prioritization based on application. Stated another way, it is "application-aware". It can perform a crisp prioritization of network traffic based on the application involved regardless of whether or not multiple applications are running on the same server.

In many situations, a "weighted fair queuing" prioritization would perform a better job of providing the balance an end-user wants.

Videoconferencing requires a substantial amount of network bandwidth in order to transmit decent quality moving images (along with voice) from one endstation to another. In addition, the real-time aspects of videoconferencing require a small and consistent latency (delay) of the information as it moves through a network. Large latency (large delay) results in time delays that make conversation awkward. Inconsistent latency results in choppy audiovisual quality.

Document sharing requires a reasonable amount of network bandwidth, since a whole screen image must be updated to multiple people. Although the real-time aspects of document sharing aren't as critical as videoconferencing, people are in fact interactively sharing and commenting on information, and thus the latency must be kept fairly reasonable.

Computer models can be extremely large, and thus moving them from one endstation to another requires very large bandwidth. If such transfers aren't accomplished quickly, end users tend to lose productivity, or at the very least grow impatient. Latency usually isn't an issue for the transfer of computer models since there are no real-time constraints.

Since analysis is often based on computer models, it often require large amounts of bandwidth to move models and associated analysis information from one endstation to another. As with computer models themselves, latency usually isn't an issue for the transfer of analysis information.

Simulations often utilize under lying computer models, and thus require large amounts of bandwidth. However, simulations also typically involve some degree of real-time nature, such as a person viewing a simulation as a manufacturing operation moves through pre-defined stages. As such, latency can be a significant networking issue for the successful implementation of simulation technology.

The various forms of manufacturing operational information can have dramatically different characteristics requiring prioritization of the underlying network traffic. If any form of online machines are being fed information, then low latency can be a requirement. Similarly, if safety or emergency information must be accessed in special circumstances, such access may have to be very fast, or significant problems (potentially even life-threatening in some extreme cases) could result. The mix of bandwidth and latency required by these widely varying types of information realistically means prioritization in networking equipment is essentially a mandatory requirement.

Fairly often, such controls do not have large bandwidth requirements, but they may have some degree of real-time characteristics which dictate prioritization of information being sent to them via a network.

Typically, sensor information require a small amount of bandwidth. But because it may have a significant amount of real-time emphasis, latency may be important.

PDM is generally a client-server operation. Depending on what it being checked-in/out, it can require very large amounts of bandwidth, or it can require very small amounts of bandwidth. The key is to make the operation sufficiently fast so as to not cause lost productivity on the part of end users. In general, latency isn't an issue for PDM operations since there isn't a real-time aspect of the applications operation.

Enterprise management applications are typically client-server. The bandwidth requirements can range from small to large, depending on the particular area of manufacturing. And although such enterprise management applications aren't truly real-time, they do need to have an underlying network able to make them "quasi-real-time" in the sense that information can be captured quickly and easily.

Because of the diversity of information accessed through the Internet, prioritization of this traffic along with other applications will increasingly be critical to its successful use. Bandwidth, latency, and in fact security are all significant issues for the Internet as viewed as an application.

As with the Internet, the diversity of information accessed through modern intranets is significant. These diverse types of traffic indicate a strong need for an application-aware traffic prioritization.

As with the Internet and intranet applications, extranets have tremendous diversity in the types of information supported. In particular, the diversity of the latency requirements of this information requires prioritization.

Video requires a large amount of bandwidth, and consistent and low latency. Voice requires a relatively small amount of bandwidth, and consistent and low latency. Data can require very large to very small amounts of bandwidth, and often does not have latency constraints.

Email typically requires low bandwidth, and does not have latency requirements as it is typically a background task.

For all these highly diverse applications to run on top of a common networking infrastructure realistically means the underlying networking equipment must be capable of prioritizing the transmission of information based on the application involved. Without this, networks can at best be "common denominators" in terms of providing basic networking services.

Ideally, network managers, working with end users and department heads, could determine the best way to prioritize the traffic for various applications within a given work area or workgroup. Then policies could be set which provide a crisp matching of applications to the prioritization levels available in the actual networking products in use in that organization.

The SSR 2000 is targeted at a "power workgroup" such as a CAD engineering team, or perhaps a circuit design team using EDA and simulation tools. It brings application-aware networking directly to the desktop.

The SSR 8600, on the other hand, is targeted at the backbone of very large enterprises. It continues the legacy of the SSR 8000, but takes the performance and port density to even higher levels, while retaining the scalability of a modular chassis.