Understanding Network Concepts

One chapter could not possibly cover everything you need to know about networking. Entire certifications built around networking and encyclopedias of information on the subject exist. This section of the chapter aims to introduce or re-familiarize you with some general networking knowledge. After understanding these concepts, you can think about how Solaris specifically needs to be configured to participate on a network.

Having said that, a network is more than one computer linked to another to share information or resources. Although the number of possibilities for configuring a network is nearly limitless, there are some common configuration themes.

Network Topologies

The term network topology generally refers to the way that the network is physically connected. Are all your computers in the same room? If so, you have a local area network (LAN). What if your computers are located in three states? That would be called a wide area network (WAN).

The two preceding examples are pretty straightforward. There is no debating that a multiplestate network is a WAN. However, in some cases, the lines between a LAN and a WAN become confusing. For example, say you have computers located in two buildings. Is that a LAN or a WAN? Well, it depends. Generally speaking, a LAN is defined as a network of computers that share a high-speed connection. Historically, a "high-speed" connection has been defined as 10 megabits per second (Mbps). However, some new long-distance networking technologies, ones that would typically be considered WAN technologies, are approaching or exceeding 10Mbps.

To answer the question of the two-building network, you need to know how the buildings are connected. Is there a network cable connecting the buildings, and is that connection running at the same speed as the rest of the network? If so, it would be considered a LAN. If the connection between the two buildings is a dial-up connection or an equally slow network connection, it could be classified as a WAN.

Network topology can also refer to the layout of your network within your building. This refers to a much smaller scale than even a LAN designation. The four major classifications of intra-network topologies are star, bus, ring, and hybrid.

On a star network, all computers are connected to a central device known as a hub. If you think about the hub being the center of the network, with all computers radiating out as arms of the star, you get the picture. Multiple hubs can be connected to one another to extend the star (which technically becomes a star-star network, but few people care to designate that deeply). Twisted-pair cabling is common on star networks, and the star is the most popular network configuration.

Bus networks do not require a central hub. All computers are connected in a serial fashion. The main cable connecting the computers is called the backbone and is often coaxial cable. Bus networks can be difficult to troubleshoot because one break in the backbone will cause the entire bus to fail. For this reason, bus networks are not very popular.

A ring network physically looks like it sounds: a ring of computers. Like a bus, a ring does not require a central hub. Generally speaking, ring networks consist of two rings. The first is the primary ring, and the second is a redundant ring, in case the primary ring fails. Ring networks use either fiber-optic or coaxial cable.

Hybrid networks are typically combinations of stars, with multiple redundant paths. That way, if one link between networks fails, another link is available for use. You'll see hybrids only on large networks. A good example of a hybrid network is the Internet.

Network Hardware

Obviously, to get on a network, you need a computer. This computer might function as a client or a server, but regardless, it's known as a host. A host is any device on the network that has an Internet Protocol (IP) address (more on TCP/IP in a bit).

On networks, clients are the most common type of computer. A client is the computer where the user sits to do their daily work. Clients request information from servers, and in comparison to servers, they don't require hardware that is as powerful.

Servers are typically locked away in a server closet or network room. Users should not have direct access to servers. A network might have hundreds of clients but only a few servers. Servers come in a variety of forms, including everything from "regular" computer cases to rack-mounted devices. The hardware in a server needs to be powerful enough to serve its clients, which in many cases means multiple processors, terabytes of disk storage space, gigabytes of memory, and multiple network cards.

Network Cards

Every computer that wants to participate on the network needs to have a network interface card (NIC) installed. NICs are also called network adapters. Network cards are built to work with one specific networking technology. In other words, if you are running an Ethernet network, you need an Ethernet NIC. Similarly, if your network is running IBM's Token Ring technology, you need a Token Ring NIC.

Some network cards have interfaces (called transceivers, because they transmit and receive) for multiple types of cable, or multiple connectors. Unless otherwise specified, a network card can use only one of its connectors at once. If the network card has connectors for twisted-pair and coaxial cabling, you can use one or the other, but not both.

As long as expansion slots are available, most computers can handle more than one network card. In fact, to speed up network access to servers, most heavily utilized servers have multiple NICs.

Connectivity Devices

If you are running a star network, which is likely, you will need a central point to plug your computers' network cables into. As stated earlier, this central device is called a hub. A variety of hubs are available. Some common classifications of hubs include passive hubs, which do not amplify the signal, and active hubs, which do amplify the signal.

Another common connectivity device is a switch. Switches look like hubs and connect multiple computers just as hubs do. However, switches are more "intelligent" than hubs, sending the network packets to the intended recipient instead of broadcasting the information out on all its ports. Because of this, switches provide for greater bandwidth than do standard hubs. Common speeds for switches and hubs are 10Mbps, 100Mbps, and 1Gbps.

Other Network Devices

If your network has multiple IP subnets, or you have another reason to logically divide your network, you will need a router. Whereas hubs and switches connect computers on a LAN, routers connect LANs to each other.

Routers are "intelligent" devices that selectively pass or block network traffic as directed. One of the major bandwidth drainers on networks is broadcast traffic. Broadcast traffic is sent from one computer and addressed to everyone else on the network. It's like getting the junk mail addressed to "Recipient" in your mailbox. The problem is, on a network with limited bandwidth, excessive broadcasts can severely hamper productivity. Routers do not pass broadcasts, making them an important tool for segmenting networks.

There are other, but less common, devices you will encounter. Bridges are like routers, but bridges direct traffic based on physical address, not logical address. This makes them a bit less flexible, and bridges do pass along broadcasts.

Gateways translate one protocol into another. If your network is running a single protocol, you will not need to worry about using a gateway.

The Logical Network

So far, this chapter has covered physical components critical to networking. There are also a number of logical components you need to understand. Logical networking components include the language that the computers will speak, as well as the rules that computers use to communicate, much like syntax in human languages. They're called logical components because they're things you cannot touch, unlike a network card or a network cable which you can physically hold.

After your computers are physically connected, rules need to be established to govern communications. To this end, various communication methods have been developed. Among them are Token Ring, Fiber Distributed Data Interface (FDDI), Asynchronous Transfer Mode (ATM), and Ethernet. Because Ethernet is the most popular LAN communications method and is supported by Solaris, it will receive all of the attention here.

Ethernet is a contention-based network access method, based on a protocol called Carrier Sense Multiple Access with Collision Detection (CSMA/CD). On an Ethernet network, a computer wishing to communicate listens to see whether there's any traffic on the network. This is called carrier sense (it's like picking up the phone and listening for a dial tone before you begin to dial). If there is no traffic on the network, the computer sends its message, in the form of packets and addressed to its intended destination, out onto the network. Ideally, the destination computer receives the packets, and there are no problems.

But now consider an example of hundreds of computers on a network. Ethernet provides for multiple access, meaning that more than one computer can be physically attached to the network at one time. With hundreds of computers, though, more than one computer might be listening for carrier sense at one given time and might send packets out at the same time. In such cases, a packet collision happens, and neither destination receives their information. CSMA/CD provides for built-in packet collision detection, meaning that the colliding packets get bounced back to the sender, who realizes that there was a collision. After waiting a random number of milliseconds, each sender will retry to send their message.

Ethernet sounds a bit barbaric in terms of access, and in reality it can be quite inefficient. But it works, and it's one of the least expensive technologies to implement. Ethernet has standards for 10Mbps and 100Mbps (fast Ethernet) communication, as well as newer standards for 1Gbps (gigabit Ethernet) and 10Gbps. In order for your network to run Ethernet, all connectivity devices (network adapters, hubs, switches, routers, and so on) must support Ethernet. Ethernet is so common that finding compatible equipment is usually not a problem.

The OSI Model

In the early 1980s, the International Organization for Standardization (ISO) developed a seven-layer networking model called Open Systems Interconnect (OSI). OSI grew out of the increasing need for standardization in networking protocols. The OSI model is a theoretical model of how networking should happen.

OSI's seven layers describe the functions that need to be taken care of by networking protocols. How the functions are taken care of is up to the designer of the protocol. For the most part, the OSI model is used as a frame of reference only. It's not a hard and fast model that everyone copies when designing network protocols. Table 10.1 shows the seven layers of OSI.

Table 10.1: The OSI Model

Layer

Number

Description

Physical

1

Describes the network hardware, as well as signaling mechanisms on network cabling.

Data Link

2

Fragments data into frames when sending it to the Physical layer. Receives and acknowledges the receipt of frames from the sender. Provides error correction of the data within the frames. Responsible for physical Media Access Control (MAC) addresses.

Network

3

Responsible for addressing and routing packets. Logical network addresses are identified at this level. Routers work at this level.

Transport

4

Provides error correction if packets needed to complete a message are missing or corrupt.

Session

5

Establishes, maintains, and disconnects communication sessions between hosts, including security measures.

Presentation

6

Responsible for data translation, compression, and format conversions.

Application

7

The applications that users use to access the network, such as Internet browser, e-mail client, or network access client (such as Telnet or FTP).

As you can see in Table 10.1, each layer is responsible for a distinct portion of the networking process. When developers program a communications protocol, however, they don't necessarily program one piece of their protocol to match one layer of OSI. For example, TCP/IP follows a five-layer model, which corresponds to the seven-layer OSI model. The top layer in the TCP/IP model covers the functionality of layers 5, 6, and 7 of OSI.

Transmission Control Protocol/Internet Protocol (TCP/IP)

The Transmission Control Protocol/Internet Protocol (TCP/IP) suite was originally developed in the late 1960s and today is the most popular networking protocol available. TCP/IP was developed in the UNIX networking world, so it should come as no surprise that it's the default protocol for Solaris.

TCP/IP is referred to as a protocol suite because a number of protocols combine to give you full network communications. If you had only the TCP and IP protocols, you'd be missing most of the picture.

The TCP/IP networking model corresponds to the OSI model, except that it's a five-layer model instead of seven layers. Here's a description of the five layers:

Hardware layer The first layer in the TCP/IP model is the Hardware layer, which corresponds to the OSI model's Physical layer. There are no protocols at this layer. However, there are standards, including Ethernet (IEEE 802.3) and Token Ring (IEEE 802.5). In many cases, you will see networking reference materials ignore this layer because there are no protocols functioning at this layer.

Note 

The Institute of Electrical and Electronics Engineers (IEEE) designed communications standards for Ethernet, Token Ring, and most other network transmission methods.

Network Interface layer The second layer takes the place of the OSI Data Link layer. Similar to the Hardware layer, it's responsible for Ethernet communications as well as the network card driver.

Internet layer The Internet layer is where the core protocol of TCP/IP, Internet Protocol (IP), works. IP is responsible for addressing and routing network packets. Included with IP at this layer are the Internet Control Message Protocol (ICMP), which is responsible for generating IP error messages, Address Resolution Protocol (ARP), which resolves logical IP addresses to physical MAC addresses for communication, and Reverse Address Resolution Protocol (RARP), which maps MAC addresses to IP addresses.

Transport layer The fourth has two protocols: the Transmission Control Protocol (TCP) and the User Datagram Protocol (UDP). TCP is responsible for providing a reliable, connection-oriented communications channel. If an error happens in communications, TCP is the one to send an error message and attempt to fix the problem. TCP guarantees delivery of packets. UDP is a connectionless protocol, meaning that it does not guarantee delivery. Because it doesn't have the overhead of guaranteeing delivery, it's faster than TCP. However, guaranteed data delivery must then be handled by the application.

Application layer The top layer of the TCP/IP model covers the functionality of layers 5, 6, and 7 of the OSI model. The vast majority of protocols in the TCP/IP protocol suite reside at this layer. They include Network File System (NFS), Telnet, File Transfer Protocol (FTP), Hypertext Transfer Protocol (HTTP), Routing Information Protocol (RIP), Simple Mail Transport Protocol (SMTP), Remote Procedure Call (RPC), and Simple Network Management Protocol (SNMP).

When most people think about TCP/IP, they think of the Internet, and that's certainly a valid link. The Internet would not exist without TCP/IP, and TCP/IP would not be nearly as popular as it is without the Internet.

All devices on a network using TCP/IP need to have an IP address. Going into great detail about how to use IP addresses could take an entire book in its own right; instead, here are a few pointers:

  • IP addresses are 32 bits long and are written as four octets in dotted decimal notation. An example is 199.104.125.33.

  • Each node (client, server, printer, and so on) on the network needs a unique IP address.

  • Each port on a router needs a separate IP address.

  • If your network is directly connected to the Internet, you must purchase unique addresses to use.

The last pointer brings up an important topic about IP addresses. If your network is not connected to the Internet or uses some sort of translator (for example, a router using Network Address Translation), you can use nearly any IP addresses you want for your network, as specified by RFC 1918. However, if your computers are going to be out on the Internet, the addresses you use must be unique on the Internet.




Solaris 9. Sun Certified System Administrator Study Guide
Solaris 9 Sun Certified System Administrator Study Guide
ISBN: 0782141811
EAN: 2147483647
Year: 2003
Pages: 194

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net