`Interconnection Basics.
`Subnetworks. Hardware
`Configurations
`
`51-20-35
`
`\./
`
`Bridges, Routers, and
`Gateways: What Is the
`Difference?
`
`Nathan J. Muller
`
`PAYOFF IDEA. Local area networks (LANs) were originally
`designed to facilitate the exchange of information within the
`relatively narrow confines of a single organization. Variations
`and improvements on this scheme led to the need to tie remote
`locations into the LAN backbone and to link different subnet-
`works. This article focuses on the interconnection devices—
`bridges, routers, and gateways—used to accomplish this task.
`
`INTRODUCTION
`
`Initially. the LAN was used to link common sets of hardware. With their
`increasing popularity among organizations. coupled with the widespread
`availability of the Transmission Control Protocol/Internet Program (TCP/lP)
`LANs began to be used for linking diverse hardware and operating systems.
`Repeaters. bridges. routers. and gateways facilitate networking: the latter three
`relieve host computers of the processing-intensive tasks of protocol conversion
`and routing infomtation to appropriate locations. And because they can be
`shared among many users. these devices contribute substantially to lowering
`the cost of networking. Because repeaters do not have as many functional
`capabilities as the other three devices. this article does not discuss them in
`great detail.
`
`BENEFITS
`
`Each interconnection device is designed to operate in conjunction with a
`different layer of the Open Systems Interconnection (OSI) reference model
`(see Exhibit
`l). which provides specific levels of network functions. When
`properly integrated with the LAN. bridges. routers. and gateways offer the fol-
`lowing advantages:
`0 Extended network reach.
`
`0 Simplified cabling.
`
`Anomach Publisher:
`© 1990 Warren, Gorham a. Lamont, Inc.
`
`Date Communlcattons Management
`u-n
`
`Page 1 of 6
`
`FIS Exhibit 1031
`
`
`
`DATA COMMUNICATIONS MANAGEMENT
`
`
`
`
`
`BRIDGES, ROUTERS, AND GATEWAYS
`
`
`
`
`
`
`
`
`
`
`
`
`Improved overall performance.
`
`
`
`Additional configuration flexibility.
`
`
`
`Enhanced security and maintenance through partitioning.
`
`
`
`
`
`Simpler network management.
`
`
`
`Reduced operating costs.
`
`
`
`
`
`
`INTERCONNECTION BASICS
`
`
`The most basic interconnection device used with LANs is the repeater. A
`
`
`
`
`
`
`
`
`
`
`
`
`repeater regenerates a signal so it can traverse a longer distance without
`
`
`
`
`
`
`
`
`
`
`
`
`becoming distorted. For this reason. the repeater is most often used to inter-
`
`
`
`
`
`
`
`
`
`
`
`
`connect LANs that are close together. typically within the same building. The
`
`
`
`
`
`
`
`
`
`
`
`
`repeater operates at the lowest level of the OSI reference model—the physical
`
`
`
`
`
`
`
`
`
`
`
`
`layer—and can be used only to link LAN s with the same protocols. It cannot
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`control or route information. and it does not have management capabilities.
`
`
`
`
`
`
`
`
`
`
`
`In addition. repeaters are segments, with the total bus length not to exceed
`
`
`
`
`
`
`
`
`
`
`
`
`
`LSOO m. Beyond that, the delay becomes too long. which causes problems
`
`
`
`
`
`
`
`
`
`
`
`
`because most networking protocols require messages to be answered within
`
`
`
`
`
`
`
`
`
`
`a set time.
`
`
`
`Bridges, which operate at the data link layer of the OS] model, interconnect
`
`
`
`
`
`
`
`
`
`
`
`
`
`LANs that have the same type of operating system. Therefore. the bridge does
`
`
`
`
`
`
`
`
`
`
`
`
`
`not have to perform protocol conversion. In this case. bridges simply look at
`
`
`
`
`
`
`
`
`
`
`
`
`
`the packet address to see where it is going. The bridge then forwards data
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`packets destined for an address beyond the local network to other networks.
`For example. a bridge can interconnect DECnet, 'ICPi‘IP, or Xerox Network
`
`
`
`
`
`
`
`
`
`
`
`Services (XNS) networks but cannot ensure that users on one network can
`
`
`
`
`
`
`
`
`
`
`
`
`communicate with users on another. That level of performance is provided by
`
`
`
`
`
`
`
`
`
`
`
`
`a router.
`
`
`A router has more intelligence capabilities than a bridge because it can
`
`
`
`
`
`
`
`
`
`
`
`handle several levels of addressing. It keeps a map of the entire network.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`including all the devices operating at or below its own protocol level. Whereas
`a bridge only checks the packet address to see if it is bound for another net-
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`work. a router looks deeper. Referring to its internetwork map,
`it examines
`the status of the different paths to the destination and chooses the best method
`
`
`
`
`
`
`
`
`
`
`
`
`
`of getting the packet to the addressee. Routers are protocol-dependent—that
`
`
`
`
`
`
`
`
`
`is, they can be used only to link LANs that have identical protocols.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`(
`
`
`
`
`
`
`
`
`
`
`Exhibit 1. Functional Abilities q'RepeaterS, Bridges, Routers, and Gateways in
`
`
`
`
`
`
`Relation to the OSI Reference Model
`
`
`
`
`
`Layer
`T
`3
`5
`4t
`3
`2
`1
`
`
`
`
`
`
`
`
`
`
`Formflcn
`
`Appicaflon
`
`Presentation
`Session
`
`Transport
`
`Network
`
`Dela Unit
`
`Physical
`
`
`
`
`Description
`
`Selects appropriate service for appicatlon
`
`
`
`
`
`Provides code conversion data reformatting
`
`
`
`
`
`Coordinates interaction bemoan md-opplicntlon processes
`
`
`
`
`
`Provides ond-to-ond data integrity and quality of service
`
`
`
`
`
`
`
`
`Switches and mites information
`
`
`
`
`Transfers units urrnrmumwnemm dill‘iephysicalllnk
`
`
`
`
`
`
`
`
`
`
`
`Transmission onto the network
`
`
`
`
`
`
`
`Page 2 of 6
`
`IW
`Device
`
`Gateway
`
`Gateway
`
`Gateway
`
`Gateway
`
`Honor
`
`Bridge
`
`Repeater
`
`
`
`
`/\
`
`A gateway operates at the highest levels of the OSI reference model. It inter-
`
`
`
`
`
`
`
`
`
`
`
`
`
`connects networks or media with different architectures by processing pro-
`
`
`
`
`
`
`
`
`
`tocols to allow a device on one type of LAN to communicate with a device
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`on another type. Therefore. a Systems Network Architecture (SNA) gateway,
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`for example. may be used to interconnect a microcomputer network to an IBM
`SNA mainframe. The gateway, then, acts both as a conduit over which com—
`
`
`
`
`
`
`
`
`
`
`
`
`puters speak and as a translator between the various protocol layers.
`
`
`
`
`
`
`
`
`
`
`
`
`Additional Interconnection Methods
`
`
`
`Not all products fit neatly into basic categories of LAN products. Many ven-
`
`
`
`
`
`
`
`
`
`
`
`
`dors have developed hybrid products that
`include functions traditionally
`
`
`
`
`
`
`
`
`
`associated with one or another of these categories. Therefore. an intelligent
`
`
`
`
`
`
`
`
`
`
`
`gateway device may include some attributes commonly associated with
`
`
`
`
`
`
`
`
`
`bridges and routers. Another device may operate as a gateway but default to
`
`
`
`
`
`
`
`
`
`
`
`
`
`operation as a bridge under certain circumstances.
`
`
`
`
`
`
`
`Single-unit bridges and routers (i.e., brouters) are really bridges that
`
`
`
`
`
`
`
`
`
`
`include some router capabilities. Depending on the protocol or packet. the
`
`
`
`
`
`
`
`
`
`
`
`data is either bridged or routed through the LAN. A combination router—
`
`
`
`
`
`
`
`
`
`
`
`terminal server is called a trouter. This device gives small work groups the
`
`
`
`
`
`
`
`
`
`
`
`
`
`ability to connect to LANs. wide area networks {WANs). modems, printers.
`
`
`
`
`
`
`
`
`
`
`
`and other microcomputers without having to buy both a terminal server and
`
`
`
`
`
`
`
`
`
`
`
`
`a router (see Exhibit 2).
`
`
`
`
`
`To further complicate matters. the terms gammy. bridge, and router are
`
`
`
`
`
`
`
`
`
`
`rapidly falling into generic use; they no longer relate strictly to LANs. Some
`
`
`
`
`
`
`
`
`
`
`
`
`vendors of simple controllers and front-end processors are calling their prod—
`
`
`
`
`
`
`
`
`
`
`ucts gateways or bridges. Interexchangc carriers are referring to their digital
`
`
`
`
`
`
`
`
`
`
`
`acceSs and cross—connect systems as gateways to public network services.
`
`
`
`
`
`
`
`
`
`
`Likewise. local exchange carriers are touting Centrex as a gateway. and mul-
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`tiplexer vendors are portraying their equipment as bridges between public and
`private networks.
`
`
`The confusion does not end there: what the Department of Defense calls
`
`
`
`
`
`
`
`
`
`
`
`
`a gateway on its Internet
`is merely a router in 051 terminology. whereas
`
`
`
`
`
`
`
`
`
`
`
`
`
`Novell‘s NetWare bridge is the equivalent of a router in 081. Nonetheless.
`
`
`
`
`
`
`
`
`
`
`
`these specialized interconnection devices are necessary. both to prevent the
`
`
`
`
`
`
`
`
`
`
`host from becoming a throughput bottleneck and to off-load previous comput-
`
`
`
`
`
`
`
`
`
`ing resources. which can then be devoted to application processing.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`BRIDGES
`
`
`Although the terms gammy and bridge are often used interchangeably.
`
`
`
`
`
`
`
`
`
`
`there is a subtle difference: a gateway connects dissimilar networks. and a
`
`
`
`
`
`
`
`
`
`
`
`
`bridge connects similar networks. A bridge may connect two or more LANs
`
`
`
`
`
`
`
`
`
`
`
`
`within the same building or LANs that are farther apart. Local bridges are
`
`
`
`
`
`
`
`
`
`
`
`
`
`capable of operating at I6M bps; remote bridges typically operate from 4.8K
`
`
`
`
`
`
`
`
`
`
`
`
`bps to [.544M bps. depending on the type of leased line used (see Exhibit 3).
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Minor routing by bridges is achieved by using a hierarchical routing feature
`
`
`
`
`
`
`
`
`
`
`
`
`that filters local data traffic without affecting local network performance. The
`
`
`
`
`
`
`
`
`
`
`
`
`Page 2 of 6
`
`
`
`SBK-bpt Dial-up UM
`
`TCP/IP Backbone
`
`x.25
`Public Data
`Network
`
`Note:
`YOPIIP harm Comm! Protocol/Internet Program
`
`
`bridge receives packets of data. scans only to the network address. and passes
`the packets to the appropriate network, where they are ultimately routed to
`the intended addressee.
`
`A self-learning bridge can filter information by monitoring the traffic on the
`networks connected to it and learning the addresses that are associated with
`each network. In this way, the bridge isolates the traffic destined to remain
`on the local segment of the network and broadcasts the rest to the other net-
`works. After initial installation. the bridge forwards all packets it receives. As
`it learns which addresses correspond with each network or subnetwork. the
`
`
`Exhibit 3. Typical local and Remote Bridge Configurations
`Noonrtr
`loosen
`Terminal
`
`
`
`Page 3 of 6
`
`BRIDGES. ROUTERS, AND GATEWAYS
`
`level of filtering is increased. The bridge can even learn the entire topology
`of the network and be able to automatically implement reconfigurations.
`Bridges are ideally suited to interconnecting similar networks in which pro-
`tocol conversion is not required. security concerns are minimal. and only
`rudimentary routing required. For example.
`in a campuslike environment.
`bridges can be used to connect each building‘s local network to the fiber-optic
`backbone. The bridge restricts local traffic to a building or cluster of buildings
`and keeps it off the superhighway. The bridge‘s filtering capabilities enable the
`network manager to restrict the types of packets that go over the bridge.
`thereby alleviating traffic bottlenecks and limiting access to certain types of
`network resources.
`
`is
`traffic unless it
`Backbone traffic does not enter a building‘s local
`addressed to a node there. The use of bridges at this level provides an effective
`way of expanding the capacity and physical reach of computer resources while
`minimizing the performance interconnection costs at higher levels.
`
`Subnetworks
`
`Many organizations find bridges useful in dividing large LANs into discrete
`subnetworks that are easier to control and manage. Bridges can be used to
`group similar devices. protocols. and transmission media into communities of
`interest. Each community of interest constitutes a subnetwork. Such partition-
`ing can yield many advantages—for example. the elimination of congestion
`and the improvement of response time for the entire network. In addition, par-
`titioning can make adding. moving. and changing devices on the network
`much easier. because only the effects of the activities on the subnetwork must
`be considered. Finally. partitioning makes problems easier to diagnose and
`isolate and enhances overall security.
`
`More sophisticated bridges suppon multiple bridging between two net-
`works. providing the capacity required for high-volume traffic and the
`processing power to implement redundant, reliable configurations. Recon-
`figurations and initial and continuous operations are often automatic. For
`example. bridges that use the spanning tree protocol (part of the IEEE 802.l
`high-level interface standard for Ethernet LANs) can facilitate the design and
`implementation of flexible. reliable networks. It allows networks of bridges to
`be instructed when to accept or reject particular messages so the data will flow
`only over specified routes. In the event of a failure on the network. the bridge
`automatically selects alternate paths. ensuring continuous network operation.
`Bridges come in handy when multiple versions of the same LAN product
`are interconnected. For example. AT&T‘s original version of StarLAN can be
`bridged to StarLAN 10.
`the company‘s product
`for users needing high-
`bandwidth (lOM bps) LANs. AT&T implement bridging through a dedicated
`bridge unit. its Information Systems Network, or a router.
`
`In addition, bridges may be used to link mixed-media and dissimilar-speed
`networks (e.g.. Ethernet and token-ring) but are capable only of moving indi-
`vidual packets between them. Such bridges operate at the media-access con—
`trol (MAC) sublayer within the data link layer of the OSI reference model.
`
`
`
`
`
`DATA COMMUNICATIONS MANAGEMENT
`
`BRIDGES. ROUTERS, AND GATEWAYS
`
`These devices respond only to the addresses of the packets on the networks
`they bridge. however. and not to the size of the packets or the speed of trans-
`mission. Thcrefore. using bridges to link Ethernet and token-ring LANs, even
`at the data link level. may present problems with reliability. Because a high-
`lcvcl protocol conversion is involved when connecting these networks, a gate-
`way is usually required.
`
`Hardware configurations
`
`including standa-
`Bridges come in a variety of hardware configurations.
`loncs. plug-ins to multiplexers, and hybrid devices that include the functions
`of routers. One advantage of standalone products is portability; because they
`are complete. self-contained units, they can be moved wherever needed with
`minimal disruption to other network components.
`
`Disadvantages include a tendency to be bulky. consumption of large quanti-
`ties of power. and the use of a disproportionate share of available space in
`equipment rooms. In addition, standalone units usually are not expandable.
`When interconnecting remote LANs. standalone bridges have another serious
`drawback: they require a dedicated connection to a remote bridge, by way of
`either a leased line or a dedicated port on a multiplexer. regardless of how
`much the line is used.
`
`In addition to the obvious advantages of reductions in cost, power consump-
`tion. and space. a bridge that is an integral part of a T] multiplexer allows
`bandwidth allocation to either LAN-to-LAN communications or to voice or
`data communications. A Tl multiplexer with card-mounted bridges can allo-
`cate between 9.6K bps and 1.5M bps of bandwidth to the point~to-point LAN
`connection. Any remaining bandwidth may be allocated to voice and data.
`
`(
`
`(
`
`The management systems of standalone bridges can graphically represent
`the network topology at a workstation, collect network performance statistics.
`receive alerts from other bridges and links, and allow the operator to remotely
`disable faulty network elements. It is also possible for network managers to
`control access, set priorities for passing data over the network. and segment
`the network for maintenance and expansion purposes.
`
`There are advantages to integrating the bridge into the multiplexer in wide
`area networking environments. The multiplexer‘s existing management system
`can monitor and collect error and use statistics from the bridge. thereby sim-
`plifying overall network management.
`In addition.
`this arrangement
`eliminates unnecessary start-up costs, because a separate bridge management
`software package typically sells for $6,000 to $10,000.
`The combination channel service-data service unit is used to link remote
`LANs at TI speeds. Because Tl provides 1.544M bps of bandwidth through-
`put, users can break the traffic jams that inevitably results when high-speed
`(c.g., 10M bps) LANs attempt to pass data to remote LANs over dial-up tele-
`phone lines equipped with relatively slow (e.g., 2.4K bps to l9.2K bps)
`modems (see Exhibit 4).
`
`The channel service unit functions as the front end of a circuit to equalize
`the received signal. filter both the transmitted and received wave-forms. and
`
`C
`
`Page 4 of 6
`
`(
`
`Exhibit 4. High-Speed Bridge Interconnection of a Wide Area Using T1
`Neonnt
`Loom
`
`WWW —
`
`(24K butclSZK boa)
`
`Nah
`MU cmwm-aumm
`
`I
`interact with the carrier‘s test facilities. The data service unit element trans-
`forms the encoded waveform from alternate mark inversion (AMI) to a stan-
`dard required buisness equipment
`interface (e.g.. RS-232 or V.35).
`In
`addition. it performs such functions as data regeneration. control signaling,
`synchronous sampling. and timing.
`
`Depending on the complexity of the network and the protocols used, a
`hybrid bridge and router unit (i.e.. a brouter) may be appropriate. Such
`devices typically house two main boards: one makes intelligent routing deci-
`sions, and the other performs the filtering function of a bridge. For example.
`the user wants to select a pure bridging operation to achieve maximum packet
`throughput. the routing functions of the hybrid device can be suppressed by
`sending the appropriate command.
`
`ROUTERS
`
`A router is similar to a bridge in that both provide filtering and bridging
`functions across the network. Routers. however. are different from bridges
`because they offer more sophisticated network management and traffic control
`capabilities. In terms of complexity. a router falls between a bridge and a
`gateway.
`
`The function of a router is to join LANs at the network layer of the OSI
`reference model. This layer has two levels: intemet and subnetwork. Because
`DECnet. for example. does not have the internet layer. its routers work at the
`subnetwork level only. Digital Equipment Corp terminal protocol. local area
`transport. cannot be routed because it does not conform to the specifications
`of the network layer. Therefore. to be routed. an application must use a pro-
`tocol that performs the functions associated with the network layer.
`
`
`
`DATA COMMUNICATIONS MANAGEMENT
`
`Each network protocol has a routing protocol built into it. Through this. the
`router accesses the addressing information and shares it with other routers and
`hosts on the network. The information the router needs to route data is built
`into the packet itself.
`
`To send packets to their destinations, a router must perform several func-
`tions. When a packet arrives at the router, it holds the packet in queue until
`it is finished handling the previous packet. The router then scans the destina-
`tion address and looks it up in its routing table. The routing table lists the vari-
`ous nodes on the network. the paths between those nodes, and how much it
`costs to transmit over these paths.
`
`If a particular node has more than one path. the router selects the one that
`is the most economical. If the packet is too large for the destination node to
`accept. the router breaks it down into a manageable frame size. This capabil-
`ity is especially important in wide area networking. in which telephone lines
`provide the link between LANs. With smaller packets. there is less chance that
`the data will be corrupted by noise on the line. Even if that occurs and a
`retransmission is necessary, the smaller packet size reduces information delay.
`
`Routing Types
`
`There are two types of routing: static and dynamic. ln static routing. the net-
`work manager must configure the routing table. Once set. the paths on the net-
`work will ncver change. This might be acceptable for a LAN confined to a
`small geographical area but
`is not practical
`for wide area networking.
`Although a static router issues an alarm when it recognizes that a link has
`gone down, it does not automatically reconfigure the routing table to reroute
`traffic. A dynamic router, on the other hand, automatically reconfigures the
`routing table and recalculates the least expensive path. Some routers even
`rcbalanec the traffic load.
`
`Because routers are protocol specific. more than one router may be needed
`to support all of an organization‘s networking needs. Some multiprotocol
`routers are capable of routing several protocols simultaneously.
`thereby
`approaching the function level gateways.
`Routers can bypass link failures and congested nodes. which is critical for
`applications that cannot tolerate unnecessary delays or prolonged outages.
`Bypasses are facilitated by the ability of routers to share information with each
`other through the OSI network layer. Bridges cannot do this because they do
`not have access to the network level through a routing protocol. Therefore,
`when one bridge gets overloaded, the others will never know about it.
`
`GATEWAYS
`
`Because organizations generally consist of specialized work groups. differ—
`ent networks may be needed to meet the requirements of different users. When
`a device performs protocol conversions
`that allow information to be
`exchanged between the various networks. it is called a gateway.
`
`As organizations become more complex. the ability to share files and com-
`municate information across diverse networks becomes necessary to improve
`
`(
`
`Page 5 of 6
`
`BRIDGES. ROUTERS. AND GATEWAYS
`
`efficiency and productivity. The need to connect dissimilar LANs may also
`arise as the result of corporate mergers or acquisitions. or it may stem from
`the desire to interconnect LANs with WANs (e.g.. packet—switched networks)
`for economical data transport over long distance. Whatever the justification
`for linking dissimilar networks. gateways are designed to do the job.
`Before gateways become widely available in the early l980s. users had to
`purchase special boards (e.g.. IRMA. from Digital Communications Associ-
`ates lnc in Alpharetta GA or Tempus-Share. from Micro Tempus Inc in Mon-
`treal). one for each microcomputer. to permit occassional access to the host.
`Such products permit micro-to-host connections but are a very expensive eon-
`nectivity solution. Because a gateway connects several microcomputers to the
`WAN. it reduces operating costs and streamlines the network (see Exhibit 5).
`
`A gateway can physically consist of a two-port card that plugs into the
`expansion slots of the microcomputer that has been designated as the server.
`The two-port configuration allows gateways to perform speed conversion. For
`example. one port on the board might provide a 64K bps connection. whereas
`the other might provide a l9.2K bps connection. Together. the two ports can
`support up to 32 concurent sessions. Some gateways even include a built-in
`packet assemble—disassembler to provide logical access. eliminating the need
`for separate units.
`
`Access to the gateway is controlled by assigning specific ports to’ certain
`microcomputers. When a microcomputer requests access to the gateway. it is
`given the port reserved for it. Because no other microcomputer can access the
`port. security is enforced. Under this scheme, each port may have access
`privileges associated with it. One port may provide access to all mainframe
`applications. for example. whereas another port may be limited to only one
`application. The problem with dedicated access is that idle ports cannot be
`
`Exhibit 5. Gateway for LAN-to-WAN Interconnection
`
`Microcomputer
`
`
`
`DATA COMMUNICATIONS MANAGEMENT
`
`BRIDGES. ROUTERS, AND GATEWAYS
`
`used by anyone else. which means that efficiency is sacrificed for the sake of
`security.
`
`When security is not an issue. gateway access may be on a contention basis,
`providing more opportunities for users to link with the mainframe or other
`network resources because users are not limited to specific gateway ports.
`Some gateways permit both shared and dedicated access. allowing some ports
`to be reserved for specific microcomputers and the rest pooled for general use.
`
`Advantage
`
`When a separate server is used as a gateway. cabling costs and installation
`time are reduced and moves and changes are easier. In fact, users can change
`the physical location of their equipment and retain their logical address on the
`network. With communications functions off-loaded from the host. valuable
`processing resources are made available for more important tasks.
`Another advantage of using gateways is simplified network management.
`Instead of monitoring the traffic of I00 microcomputers on the network.
`managers would have to monitor only a single gateway. which appears to the
`host as a single peripheral device.
`In this case. a separate cluster controller
`is unnecessary because the gateway replaces it.
`
`the data
`Gateways are capable of extracting detailed information about
`traffic that passes through them and about that status of the data links it inter-
`faces with. The gateway can ensure that the links are handling data reliably.
`without exceeding user-defined error rate thresholds. In addition. it can moni-
`tor the various protocols being used, making sure that enough protocol con-
`version processing power is available for any given application. The gateways
`management system can generate a variety of reports that can be output auto—
`matically at specified times or as demand warrants using a keyboard com-
`mand. Network statistics can be archived for trend analysis, which can assist
`in long-range planning.
`In wide area networking environments. the gateway balances load levels.
`bypasses failed links, and finds the most economical route. With some gate-
`ways, all of these functions are performed automatically as the result of a
`single connection request from a user. regardless of the equipment location
`or the protocols involved. In this environment. the ability to detect, isolate.
`and diagnose problems becomes important. The network management tools
`that are available with today‘s sophisticated gateways allow the remote config-
`uration of channels.
`links. and such network interconnection devices as
`bridges and routers. Through the network management system, gateway ports
`may be brought online or offline as required.
`
`Because gateways perfonn protocol conversion at every layer of the OS]
`reference model, performance bottlenecks may become a problem. Every new
`connection, hop. and protocol that is added to the network not only intensifies
`the problem but invites new problems (e.g.. higher system costs,
`limited
`growth and expansion. and nontransparent connection), which complicates
`network management. With so much networking overhead devoted to protocol
`translation, gateways have been relegated to such specific applications as elec-
`tronic mail and batch file transfers.
`
`Page 6 of 6
`
`Intelligent Gateways
`
`Some vendors are developing so-called intelligent gateways that communi-
`cate with each other to determine the best way to route information, taking
`into consideration such factors as congestion. priority. performance (i.e..
`throughput. delay. and error rate), security. and even cost. Building such
`capabilities into intelligent gateways relieves users of having to make these
`decisions.
`
`An inherent weakness of such schemes, however. is congestion, which may
`affect the performance of the entire network or of only one gateway of the net-
`work. Congestion may be caused by an inefficient routing scheme, creating
`traffic to stay on the primary data link longer than necessary. which slows
`down the entire network. Alternatively. congestion may actually be in the
`gateway itself, which occasionally occurs when there are too many packets to
`filter. To minimize the chance of such bottlenecks. the gateway protocols must
`be able to perform flow control and respond to congestion indicators. Inves-
`tigating the cause of the congestion can make a difference in determining
`whether it
`is worthwhile to reroute through other gateways attached to the
`network.
`
`When congestion is detected, the intelligent gateway can assign priority to
`the information that is to be routed (e.g.,
`it can determine whether local or
`internetwork traffic should be given preferential treatment). Ranking informa-
`tion is also important in the management of the network. Intelligent gateways
`have the ability to let diagnostic information pass through or around congested
`areas. providing real-time status reports on each link.
`
`If the entire network is congested, the packets can bypass the alternative
`gateways located on the other side of the network in favor of hopping through
`a different network. Hopping to another network may pose a threat to security.
`however. Intelligent gateways maintain security by distinguishing between rou-
`tine and sensitive information during the routing decision.
`
`CONCLUSION
`
`Bridges. routers, and gateways provide varying levels of connectivity. effi-
`ciency, and economy to corporate networking. The choice of interconnection
`device hinges largely on the topology of the network and the types of applica-
`tions being run on the network. Such devices are increasingly appreciated not
`only for providing internetworking connectivity but for helping to unify the
`organization into an enterprisewide utility.
`
`A former consultant. Nathan J. Muller now heads the Consultant Relations
`Program at General DataComm inc in Middlebury CT. He has 18 years of expe-
`rlence in the computer and telecommunications industries and has an MA
`degree in social and organizational behavior from George Washington Univer-
`sity. Muller has written extensively on all aspects of computers and
`communications and is the author of Minimum Risk Strategy for Acquiring
`Communications Equipment and Services (Norwood MA: Artech House, 1989).
`
`