Device Application
Bridge Connects multiple network segments and forwards data between them Content Filter Prevents access of restricted external Web content
Firewall Prevents unauthorized users from accessing the network Gateway Links two networks that use different protocols
Hub Provides a central connection point for a network configured in a star topology Modem Connects a workstation or LAN to an outside workstation or network, such as the
Internet
Monitor Monitors activity on the network by node and by network segment
Router Sends data transmissions only to the portion of a network meant to receive them Satellite Transmits signals from a server in orbit
Switch Selects network paths at high speeds
UPS Provides uninterruptible power for network electronics, especially servers
Wireless Hub Provides mobile, cable-free access to servers, shared resources, and the Internet from anywhere within range of the hub
Wireless Modem Allows workstations and laptops to communicate with a wireless hub (access point)
Servers
The centerpiece of most bioinformatics networks is a server (or more than one) that supplies files and applications to workstations, printers, and other clients. Servers are typically high-speed
dedicated computers with several GB of RAM, multi-GB fast hard drives, and over-engineered power supplies that can withstand power surges and other challenges. Servers vary in size and shape, degree of redundancy, performance, expansion options, amount of noise generated in normal operation, the type of operating system supported, management software, security features, power supply design, amount of cache memory, and price.
Servers are no longer relegated to footlocker-sized cases, but are available in units as small as a pizza box that can be easily stacked in racks to provide high server densities. Related to form factor is the operating environment, in that the compact size often necessitates the use of high-volume fans that not only move large quantities of air over the densely populated motherboards, but that
generate considerable noise as well. As such, servers may need to be mounted in a separate room or closet, away from researchers whose work the noise may disrupt. Also related to form factor is the provision for redundancy in the two most common server failure points—the mechanical disks and the power supplies. Many server designs provide internal redundant disks and power supplies that take over as soon as the main units fail.
The typical server used in a bioinformatics laboratory has between 1 and 8 GB of RAM, several hundred GB of disk storage distributed between 2 and 8 drives, 2 power supplies, BIOS password protection, and virus protection. Performance, as measured by throughput in Mbps average response time in milliseconds, and thousands of requests handled per second, is a function of the processor, operating system, amount of RAM available, cache memory, and overall design.
The most common server operating systems are Microsoft Windows 2000, Linux, Solaris, UNIX, and Microsoft .NET. Windows 2000 commands about a third of the server market, in part because of the familiar graphical user interface (GUI) and compatibility with relatively inexpensive server hardware. The relatively new Microsoft .NET Server is Windows 2000–based with added Web development tools. Linux, an increasingly popular operating system for servers and bioinformatics workstations, accounts for only about 5 percent of the overall server operating system market. An advantage of using Linux as a server operating system is cost savings and an abundance of license-free (albeit Spartan)
utilities. Linux is considered more stable and reliable than Windows 2000, but more difficult to use. In comparison, Solaris commands a little over 15 percent of the server market, followed by IBM AIX and HP's UX. These various flavors of UNIX account for over a third of the server market, especially in high-end applications, such as massive sequence databases.
In addition to generic servers that serve content to clients on the network, there are specialized server designs, such as cache, file, print, mail, proxy, and terminal servers. A cache server
dynamically pulls frequently accessed content from the main servers and maintains the content in cache for later use. The purpose of a cache server is to speed content to clients and to reduce network traffic at the server site. One of the challenges with cache servers is ensuring that the cached files are current and synchronized with the files on the source server. Cache servers usually double as proxy servers, which are designed to intercept and manage client requests in a way that provides increased security by matching incoming messages with outgoing requests. A proxy server acts as a filter that passes valid requests on to a file or Web server or, if it's configured as a cache server, serves the content from its cache. Because the functionality of proxy, firewall, and caching servers is so tightly integrated, they are commonly combined in a single device.
A file server is a server configured to allow workstation clients on the network to use the disk storage on the server for collaborative work, to facilitate archiving, and to provide additional disk storage. File servers typically contain large, high-speed hard drives and comprehensive data management
software. Print servers provide buffering and queuing for networked printers.
Web servers provide HTML pages or files to a Web client. A mail server hosts the e-mail system for users on the network, providing processing and storage for e-mail messages. Terminal servers connect several terminals, including dial-up modems, to a single LAN connection. A terminal server has a single network interface and several ports for terminal connections, allowing several terminals to be connected to the network by a single LAN cable.
Remote access servers, also known as communications servers, provide access to users seeking to use a network remotely, especially while traveling away from the main office. A remote access server is typically configured with a firewall and a router to provide security and to limit the remote access to a specific subset of the network. For example, a remote access server may allow access to e-mail and non-confidential files. In this way, if a hacker manages to somehow gain access to the network through the remote access server, he won't be able to destroy or steal confidential data. A remote access server is typically configured with one or more telephone modems so that remote users can call in to the network and read their e-mail and access files from any location with telephone access.
Bridges
A bridge connects two or more network segments and forwards packets between them, amplifying the signal to compensate for the loss associated with splitting a signal across multiple segments. So- called dumb bridges are protocol-specific and are designed to connect networks running the same protocol. These devices simply accept data packets from one segment of a network and forward them on to the other segments. They have no built-in intelligence.
In addition to these bridges, several varieties of bridge design provide processing, enabling data sharing between otherwise incompatible networks. For example, encapsulating bridges encapsulate network data with header information so they are compatible with devices in the destination network. A translating bridge goes one step further and actually translates the data from the source network so that the protocol is compatible with that of the destination network. A filtering bridge, also called a multi-port bridge, directs data from the source network to a specific segment of the destination network, thereby reducing unnecessary traffic on some segments of the network. In addition, there are numerous bridge designs that combine filtering, routing, and security functions.
Routers
A router directs data to the portion of a network meant to receive it rather than broadcasting data to every node of a network. Instead of merely passing information like a dumb bridge, routers monitor network activity and change traffic patterns if necessary to maintain efficiency or throughput. Intelligent routers dynamically reconfigure the communications path to improve availability and reliability.
Routers are rarely used alone but are combined with other devices. For example, routers are located at every gateway and are often included as part of a network switch. Routers are also commonly combined with a network bridge in the form of a brouter. In contrast to switches, routers are typically used at the edges of a network, where intelligence is needed to determine the best path for data.
Switches
A switch is a device that selects a circuit for sending data through a network. A switch, which tends to be simpler, faster, and less expensive than a router, lacks information about the network that a router may use in determining the best circuit or path to use to move data from one part of a
network to another. Switches, which lack the intelligence of a router, are normally used in the network backbone and at gateways, where speed is of the essence. Also called LAN switches, data switches, and packet switches, they typically contain buffer memory to hold packets briefly until network resources become available.
Gateways
A gateway links two networks running different protocols by functioning as a router and providing translation and amplification of network signals. Because gateways can connect networks using different protocols, they are slower than simple routers.
Hubs
A standard wired hub is the center of a network physically connected in a star configuration. These hubs generally have little intelligence and serve primarily as a common connection point. However, hubs can also be complex devices that provide bridging and routing between multiple LAN
architectures.
Wireless hubs, also known as access points, function like wired hubs but use different protocols that provide for different levels of interoperability. With a wireless hub, a wireless LAN can be established quickly with only a server and wireless modem cards (or PCMCIA cards for laptops). Except for the wired connection to the Internet, there is no need to drill holes in walls and pull cables to individual workstations.