Networks: The Hub of Connection
By Lynda Nolen, Bishop & Associates Inc.

More than likely, if you are reading ConnectorSupplier.com, you are using a network. In fact, the majority of us spend a great deal of time using one or more networks every day. We communicate with co-workers and friends, share files, check inventory, surf the Internet, and download drawings and specifications. For many of us, a network provides the tools necessary for us to perform our jobs. A great deal of time and effort has gone into explaining the different types of networks and how these networks work and interact. There are LANs, or Local Area Networks; MANs, or Metropolitan Area Networks; WANs, Wide Area Networks; as well as WLANs, or Wireless Local Area Networks; and SANs, or Storage Area Networks.

A network is composed of four main elements. The first are the nodes. Nodes include things like computers, printers, scanners, and PDAs; basically anything connected via wire or wireless to the network. The second is networking equipment, which includes equipment like routers, hubs and switches, servers, network cards, storage systems, and transceivers. The third is protocol, or the common language used to enable the nodes and networking equipment to communicate with each other. The most common protocol used today is Ethernet; others include InifiniBand, Fibre Channel, Fast Ethernet, and Fiber Distributed Data Interface (FDDI). The fourth component is the connection media. The connection media can be wired—with twisted pair, coaxial, or fiber optic cable, or wireless—using wireless access points. It is the connection media that links the computer or other peripheral devices to the networking equipment and allows the information, governed by the protocol, to flow from one node to another. Without proper connection media, whether it is wireless or wired, a network cannot perform correctly.

Connection Media and Networking Equipment
Routers—A router joins one network to another network. For instance, a router can join a LAN network to a WAN network, or join two or more LAN networks to a WAN network. Routers filter incoming and outgoing information—based on the addresses of senders or receivers—and determines the best path for information to travel. If necessary, a router can provide the translation of different protocols from one network to another. Routers generally have a hardwire firewall that prevents unwanted information from flowing back through the network. There are generally two types of routers, traditional routers used by medium and large-scale enterprises, or carrier and service provider networks and broadband or home routers used by consumers or small businesses. Routers can be wired or wireless, and incorporate a variety of interconnects including RJ-45, SFP, RF connectors, and fiber optic connectors.

Traditional routers are most often a variation of a standard model, built with a specific customer’s requirements in mind. These routers are generally expandable and upgradeable through the use of port adaptors and modules. They are available in a variety of sizes, from one-slot chassis all the way to 13-slot chassis. Some of the newest routers, Integrated Service Routers, not only join two or more networks, but also offer voice, video, data, and Internet access. These sophisticated routers also increase network security by transferring the responsibility of security from individual computers and users to the network itself.

A variety of interconnects are used in traditional data center routers, including SFP ports as illustrated above and RJ-45s. For routers that incorporate RJ-45s stacked and ganged jacks with integrated magnetics, such as those offered by Molex, make the designer’s job easier. As explained by Molex Market Development Manager Dave Brearley, “all Ethernet circuits need magnetics in the signal path. The designer can place the magnetic packages directly on the PCB, but they occupy space and require pick-and-place for each device. The signal path from the magnetics to the jack becomes an analog circuit, with different isolation and design rules than those for the digital circuits found on the rest of the board. By containing the analog portion of the circuitry inside an integrated connector, the designer’s job is all in the familiar digital design space. Integrated jacks also save space and reduce component count.” Presently jacks with integrated magnetics are more costly than standard RJ-45 jacks, but as more manufacturers enter the market, and higher volumes are used in applications where space is at a premium, prices are expected to be lower.

Broadband or home routers often combine the functions of a network switch and a firewall in a single unit. Wireless broadband routers, such as D-Link’s RangeBooster N 650 Wireless Router, utilize three external reversed SMA connectors for antenna connections, four 10/100 RJ-45 LAN Ports, one four-pin USB Type A management port, and one 10/100 RJ-45 WAN port. An Ethernet broadband router, such as Motorola’s BR700, utilize four-port 10/100Base-T (RJ-45) Ethernet LAN connections, and one 10/100Base-T (RJ-45) WAN port.

Based on the size and configuration of the router, a variety of additional ports can require interconnects, including serial ports (synchronous/asynchronous), SIC (Smart Interface Card), MIM (Multi-function Interface Module) slot ports, enhanced service module slots, voice co-processor module slots, multiple USB ports, and, of course, a WAN port. These additional ports all require interconnects that have the capability to reliably transmit data, whether it be voice, data, or video.

Hubs or Switches—A hub or switch is used to connect multiple devices within a LAN or a SAN. A hub, which is predominately used in small or home networking systems, unlike a router, doesn’t read any of the information passing through it or care where the information is going. A hub just takes the information in, verifies the strength of the signal, possibly amplifying it, and sends it out to all the other devices on the network. Generally there are three types of hubs; passive hubs that just send information along; active hubs that amplify the signal if necessary; and intelligent hubs that offer all the features of an active hub, but are also expandable, allowing multiple units to be stacked on top of one another. An intelligent hub typically offers remote management capabilities.

All hubs must be connected to all the necessary nodes within the network. The connection is generally made through a cabled or wireless device that runs from the hub to the NIC (network interface card) that is installed on the device. Wired interconnections used on hubs include USB connectors and IEEE1394 (FireWire) connectors.

A switch is similar to a hub in that it transmits data from one node to another node, but unlike a hub that isn’t interested in who the information is addressed to and transmits it automatically to all the other nodes on a network, a switch after determining the source, reads the address and only sends the information to the intended receiver, eliminating bandwidth sharing. Switches, unlike hubs, also have the capability of establishing a temporary connection, which is terminated once the information is sent or received, allowing increased transmission rates and virtually no collision of data.

The most common type of network switch used today is the Ethernet network switch, designed to support 10Mbps, 100Mbps, or 10/100 Mbps. Molex’s Dave Brearley commented, “Most new installations are installing one Gbps RJ-45 Ethernet connection because the premium for Gigabit has become quite low. For example, planners of new LANs generally will install the cable suitable for the next generation of speed, figuring that their network equipment can be upgraded in the future without further changes to cable plant.” The number of ports for connecting devices is based on the model of the unit, with consumer-grade products traditionally supporting four or eight ports.  In business enterprise applications, switches are often daisy-chained to support larger numbers of devices on the same LAN.

Servers—Although it may not always look like a standard computer, a server is just that, a computer that provides a service or resource to another computer. There are many types of servers; file servers which act like computers, storing files for easy access; print servers that manage a group of printers or a group of computers connected to a particular printer; a network server that directs traffic across a network; and a database server that processes inquiries, not to mention email servers, web servers, or proxy servers. The type of server used depends primarily on its application, how much data will be stored and retrieved, the number of user requests expected, and the number of users or nodes that will be accessing the system.

Presently the vast majority of servers act as client/server networks, meaning particular clients or nodes, such as printers, computers or faxes, are all hooked up to a particular server. In a client/server network, the server acts as a centralized location or computer to store or manage information for each of these nodes. Unlike a computer, which can perform many of the same functions as a server, a server is dedicated to performing just a particular function or set of functions. This is especially important when you are talking about the ability to store large numbers of files and the ability to have multiple clients accessing these files at similar times. Often in situations like this, where the loss of one server due to scheduled or unscheduled downtime can severely affect a business or organization, server clustering is initiated to provide redundancy and failover protection.

In server clustering, two or more servers are linked together in a virtually invisible manner. Server clustering allows for increased scalability, greater network flexibility, and protects against service interruptions from hardware failure, human error, or maliciousness. By using clustering, businesses make data and applications available on multiple servers linked together in a cluster configuration. Using a variety of software, these server nodes can be connected to a storage server, allowing an “everything shared” environment. In these cases, it is not only the software used, but the configuration of the server, additional memory, hard drive capacity, and also the network connections between the server nodes, that have to be considered in the creation of this flexible, multipurpose invisible environment. In order to achieve the low-latency, high-bandwidth, and high-CPU utilization needed to attain the highest performance from a server cluster, many customers turn to interconnects like FCI’s EyeMax® High-Speed Serial I/O Links for Infiniband. Based on data transfer rate and cable length requirements, EyeMax® cable assembly technology can deliver aggregate bandwidth capabilities that range from 10Gb/s (SDR) to 20 Gb/s (DDR), providing a low-cost alternative to fiber optics.

FCI offers the EyeMax® system in both pre-terminated, customer-specified cable assemblies, or as a plug kit designed for easy customer assembly. As discussed with Jim David, global marketing manager for cable assemblies at FCI, there are advantages to buying the completed assembly from FCI over building the assembly yourself or having a cable house build it for you. “As speeds increase, the sensitivity to variations in component and cable geometries and the need for equalization in cable assemblies needs to be carefully considered. By providing a full link offering, FCI qualifies the complete cable link, thus assuring signal integrity from transmitter to receiver.” In addition to supplying cable assemblies and plug kits, FCI also offers latching or thumbscrew versions of the corresponding board-mount receptacles with optional alignment pegs for PCB thicknesses from 1.6mm to 4.0mm. In addition to enabling InfiniBand cable links, EyeMax® cable assembly and connector products have also been selected as the external I/O copper link for 10G Fibre Channel, 10G Base-CX4 Ethernet, and SFF-8470 for Serial ATA (SATA) or Serial Attached SCSI (SAS).

This demand for increased speed and bandwidth doesn’t appear to be going away anytime soon, especially with companies like Microsoft installing an average of 10,000 additional servers a month. David Sideck, global market manager for high-speed and power products at FCI, says, “With continued growth in the Internet and IP traffic, particularly as consumers and businesses embrace video content, with growing numbers of users and access points, with increased server bandwidth demands in the data center resulting from CPU advances, increased bus speeds and memory performance, and with sharing of I/O resources by blade servers, today’s Gigabit cable links are evolving to 10 Gb/s links, and the need to aggregate those individual links will drive demand for 40 Gb/s and even 100Gb/s bandwidth for network uplinks.”

In a white paper published in June by Cisco, entitled Cisco Visual Networking Index–Forecast and methodology, 2007-2012,“consumer and business Internet Protocol network trends show IP traffic to increase at a combined annual growth rate (CAGR) of 46 percent from 2007 to 2012, nearly doubling each year.” This will result in an annual bandwidth demand on the world's IP networks of approximately 522 exabytes2, or more than half a zettabyte.”

And as mentioned, what is creating this demand is the implementation of applications like video on demand (VoD). Unlike other types of video, such as satellite video, digital terrestrial television video, or even most gaming applications (that create a “one too many” atmosphere), VoD requires a new stream each time a client logs in and, with each of these streams, large amounts of data must be transmitted over a network. By 2012, as indicated by Cisco, “Internet video will account for 50 percent of all consumer Internet traffic,” with the largest percentage being video-to-PC.

So, what is happening to enable this requirement for greater speed and bandwidth? According to Molex’s Brearley, “State-of-the-art networks are quickly moving toward 10Gbps. Today, this is most commonly done with SFP+ packaging and fiber connectors, because of their lower power consumption and heat generation, or with copper jumper cables.”

The semiconductor industry is working to develop 10G base T capable chips that will transmit up to 10Gbps over twisted pair copper, and research is ongoing to work with the power issues created with 10G base T chips that tend to be over four watts per port.
 

A trend is in place to pack more I/O onto a single card. Semiconductor technology is enabling more channels in fewer chips, which is passing the bottleneck to the front panel of the router. One solution Molex offers is to use iPass cables, which double the density to the front panel. Molex has also developed a patch panel using cassettes with 12 RJ-45 ports on the front, and two cable connections on the rear to jumper to the router. Current patch panels normally use punch down blocks on each RJ-45 port. The installer must terminate individual conductors in the field. Using the new iPass patch panels, pre-terminated cables are just plugged into cassettes, dramatically reducing time needed for installation and test. This approach uses 100 percent factory-terminated and tested cables, reducing installation labor by as much as 80 percent, compared to discrete patch cables. The cassettes can be either copper or fiber or both in a 1U panel.

Will the patch panels and connectors on routers for 10Gb/s look similar to routers today? According to Molex, yes, they will, but with enhanced internal magnetics to handle the higher speeds and provide acceptable crosstalk levels. They also note “CAT6A cables for 10G base T will have to handle 500 MHZ frequencies compared to 100 MHZ for today’s 1GB/S systems. This requires improvement in semiconductors, jack magnetic designs, and cables. Up to this point, most 10G base T solutions have been proprietary solutions from a specific vendor. Now, some of the individual components (semiconductors, jacks and cables) are beginning interoperability testing.”

Enhanced features, broader areas of coverage, and faster speeds will all work to create better connections for us in the future.

Lynda Nolen Product Specialist, Bishop & Associates Inc. Lynda Nolen has been in the interconnect industry for over 28 years. She has worked in sales, sales management, marketing, and product management for such companies as TRW Electronics Components Group, Sunbelt Components, Cinch Connectors, Arrow Electronics, PEI Genesis, and Delphi Interconnect. Nolen has extensive experience in competitive cross-referencing, drawing, web and catalog review, new product introduction programs, harness and connector assembly programs, account management, and customer service programs. Lynda received her Bachelor of Arts degree from Roger Williams University in Rhode Island in 1979, and has completed various electrical engineering courses.