Switches





Welcome to Jenny's Switches Page





Congested Networks

Today's Networks are becoming congested and sluggish because they weren't designed to handle the extreme amount of inter-subnet traffic currently crossing the enterprise.

Ethernet switching devices increase overall network performance by decreasing the amount of extraneous traffic on the individual network segments attached to the switch. When the packet arrives at the switch the packet's destination address is inspected to determine which segment it is destined for. If it is destined for a node residing on a different segment than that on which it arrived, the packet is then forwarded to that destination segment only.

Switches beging monitoring network traffic immediately upon powerup and automatically filters packet by the source and destination addresses as soon as a network map can be constructed (a matter of a few minutes). Additionally, all bad or misaligned packets are filtered anytime they are found.







Front panel of a switch





Toward a New Architecture

Adopting switching technology to solve bandwidth, scalability, and manageability problems can be as easy or as difficult as the network planner, makes it. Switches can be seamlessly slipped into the network to fix a trouble spot - or the network may be re-designed around switching as the core technology. Although using switching as a point-solution is often an easy win, this approach misses an opportunity to take advantage of switching technology on an enterprise scale. In order to get maximum value for an IT dollar, network owners must plan the evolution to a switched enterprise network. To do this the Network Planner must ask these questions. What type of switching should be used? Where should it be deployed? What should you do to prepare to adopt switching? How much planning is needed?



Developing a Network Architecture

Planning means developing a network architecture. An architecture is more than a laundry list of the boxes and topologies scattered throughout closets, floors and buildings--more than routers and hubs and star topologies. An architecture sets forth both the strategies and blueprints which define major network elements and the relationship between them. There is either a documented, well-conceived design, or there isn't. And the difference between an enterprise infrastructure and meshes of loosely connected networks is an architecture plan.

Architecture provides the high level design, specific subnetwork technologies, network concentration, switching and interconnection schemes, carrier services, wiring, management systems, and protocols from which implementation flows. Architecture establishes the key points of stability in networks through which next generation technology is integrated into the installed base of systems. The rigorous definition of these interfaces permits operational flexibility. An architectural framework for the network allows organizations to cope with rapid unforeseen changes in technology or business practices.



Switching Technology Is Rapidly Changing Networking

The networking industry--and by extension, network owners--are caught in a period of radical change right now. Switching technology in general, and ATM in particular, represents a discontinuous (rather than continuous, or incremental) change in networking. These technologies are not just somewhat faster than existing networks--they are two orders of magnitude greater. Such large, discontinuous change is defining. With the arrival of switching, networking has evolved into something entirely new. Although switching technology can be made to behave like the installed base internetworking devices, its power, and the compelling reasons to adopt switching, come from its ability to transform the way networks are designed, built, and run. Successful network planners will get out in front of the wave of change and begin planning now to adopt switching technology.




Picture of a Switch





Fitting The Pieces Together

Many roads lead to the enterprise switched network. This is necessary, largely because of the variability in networking technology across companies. Fundamentally, the high level structure of a switched network is simple:

A high speed point-to-point switched core providing at least 100 Mbps links to aggregate and forward traffic across an enterprise network;

High-speed frame switching (either Ethernet or Token Ring) providing dedicated links to desktops, servers, and workgroup-level resources; and Mechanisms which integrate existing broadcast-based, shared access LANs into the heart of the switched environment.



What Is A Switched Enterprise Network?

This figure provides a conceptual overview of how the various switching technologies fit together to form an enterprise network.


Anatomy of an Enterpriser Network



The core of the network (the backbone and select workgroups) is driven by high speed ATM connections. Attached to the ATM core are devices and workgroups connected via switched Ethernet and switched Token Ring. The world of legacy shared access LANs is in turn connected to the switched infrastructure through traditional bridges and routers.



A Fine Sauce; Not Dessert

The capable enterprise network must successfully combine an array of technology, old and new--including routers, frame switches, shared access Ethernet and Token Ring, bridges, cell switches, VLANs, hubs, NICs, protocols and repeaters. The resulting network architecture is not defined by a single element. Instead, it is rather like a fine sauce, blending many elements into a creation better than any of its ingredients. Individual technologies are combined in a complementary manner, each leveraging other building blocks in the network.



Store and Forward versus Cut-through Design

To decrease the amount of time to process packets, some switches use a procedure known as a "cut-through". Only the first few bytes of the Ethernet packet are read for the destination and source addresses; these switches then route the packets ahead without examining the remaining bytes for any possible problems. This procedure allows for fast processing, but allows misaligned or packets being jammed because of late collisions to be forwarded on to other network segments.

The "store and forward" approach examines the entire packet. Devices using the store and forward technique filter out any bad or misaligned packets and then filter or forward the remaining valid packets to their correct destinations. The store and forward approach offers a solution which is very "network-safe" and allows for additional features such as selective filtering schemes. Store and forward technology is now nearly as fast as cut-through designs, with the added benefit of bad packet filtering. The LSB4 combines these features to offer high performance switching between Ethernet segments with maximum filtering options and Spanning Tree Algorithm Support.



Should You Build a Switched Enterprise Network?

A switched enterprise network -- built around ATM, switched LANs, and VLANs -- offers the broad set of capabilities required to meet the demands of a new generation of application technology and organizations pushing that technology to the limit. But what are the decision criteria that indicate whether you should consider switching technology as the foundation for your enterprise network? Any enterprise that meets three of the following six criteria should begin a serious evaluation of switching technology.

Consider switching if (the):

Number of end stations in your enterprise is growing at the rate of 25 percent per year. In expanding the network, switching technology must be considered. Frame switching may be adopted at the floor or building backbone level, perhaps in place of workgroup class routers. Existing power users and servers might be connected with switched Ethernet or Token Ring, passing down existing repeated network connections to new users less likely to need high-bandwidth links.

Client/server computing applications are a success. The next killer application is probably no further away than the next scheduled client/server application rollout. Although client/server is a faster, cheaper environment for developing applications, these systems require at least an order of magnitude greater bandwidth than do terminal/ host applications.

Building backbones need more than 100 Mbps. When FDDI runs out of gas, it is time to consider switched 100 Mbps Fast Ethernet or ATM. Think about switching technology before segmenting that FDDI backbone again.

Some servers need more than 20 Mbps (shared) bandwidth. When servers with multiple shared Ethernet or Token Ring interfaces cannot support aggregate workgroup or server bandwidth demands, it is time to begin investigating 100 Mbps Fast Ethernet or ATM interfaces.

Some desktops need more than shared 10 Mbps. When advanced applications increase bandwidth requirements, it is time to consider switched LANs for high-performance workgroups. In the extreme case, for users of scientific visualization applications, or users who live in the world of multimedia or delay-sensitive applications (options trading, for example), ATM interfaces will replace existing or planned FDDI network adapters.

Servers are being centralized, increasing the amount of traffic that traverses building or campus backbones. Experience shows that as soon as more than about one-third of client/ server traffic leaves an individual workgroup, the workgroup link to the backbone must be doubled in capacity.



Incorporating Switches into your network





Switching Technology For Everyman

Switched networks offer something for everyone. New twists on venerable Ethernet provide dedicated 10 Mbps connections to every endstation on the network--users and servers alike. Links to critical shared resources are built with Ethernet delivering 100 Mbps. These links aggregate traffic streams from many desktops to workgroup file or database servers. Switched Token Ring is also making an appearance, alleviating congestion in the largest ring networks while working with the installed base of NIC cards, just as switched Ethernet does today.

Switching technology and products offer network planners the opportunity to solve networking problems with a better set of tools. Switching promises higher performance, greater scalability, and improved manageability. To capitalize on these benefits, network owners must evaluate the existing network, determine where the troublespots lay, and decide how and where to use new technology to overcome existing weaknesses. Answering these questions is the heart of the architecture planning exercise. Good answers are not easy to come by, but are proof that planning pays.



THE END


Any questions?

© 1997

Jn_White@venus.nmhu.edu