Transport Protocols
You
may have learned by now, that designing network protocols usually is done in
pieces, with each piece solving a small part of the overall problem. By convention, these protocols are regarded as
layers of an overall set of protocols, called a protocol suite or a protocol
stack.
This following learning material examines a variety of actual protocols and protocol suites, such as TCP/IP, IPX/SPX, NetBEUI, AppleTalk, and DLC.
This part covers the Planning section of the
Networking Essentials exam:
Objectives: For students to select the appropriate
network and transport protocols for various Token Ring and Ethernet
networks. Protocols include the
following:
·
DLC
·
AppleTalk
·
IPX
·
TCP/IP
·
NFS
·
SMB
1. NetBEUI
operates at the _____________________ protocol levels.
A. Application
and Presentation
B. Data
Link and Physical
C. Transport
and Network
D. Session
and Transport
2. UDP is part of the _____________ protocol
suite.
A. TCP/IP
B. IPX/SPX
C. AppleTalk
D. NetBEUI
3. TCP/IP
is __________________ than NetBEUI.
A. faster
B. slower
C. easier
to install and configure
D. none of
the above
The OSI reference model is a standard describing the
activities at each level of a protocol stack.
The OSI reference model is useful as a conceptual tool for understanding
protocol layering. Although some
protocols have been designed in strict conformance with the OSI reference
model, full OSI compliance hasn't become popular. The main influence of the OSI reference model is as a conceptual
framework for understanding network communication and comparing various types
of protocols.
Protocols are real implementations of the conceptual
rules defined in the OSI reference model.
Some protocols and protocol suites existed before the OSI reference
model was published and can be matched only loosely to the seven4ayer model.
Packets
and Protocols
Before investigating protocols and protocol stacks,
take a moment to quickly review some of the protocol-related issues discussed
in previous chapters.
The purpose of a network is to exchange information
among computers, and protocols are the rules by which computers
communicate. Computers, like humans,
can adopt any number of systems for passing messages as long as the sending and
receiving computers are using the same (or compatible) rules. Computers, therefore, must agree on common
protocols before they can communicate-failing to do so would create a
bewildering situation similar to what you'd face if you read a book in Russian
to a listener who speaks only Cherokee.
The
NDIS and ODI standards greatly simplify the task of finding common
protocols. NDIS and ODI enable several
protocols to operate simultaneously through the same network adapter card.
You can classify the many tasks that network
protocols must oversee into a few basic categories. Think of these categories chronologically, as a series of steps
(each step including a collection of related tasks) that must take place before
the data can reach the transmission medium.
These steps are the layers of a protocol stack. In one sense, the term layers more than
metaphorical.
Each layer of the stack (the Application layer, the
Presentation layer, and so on) adds a layer of information to the packet, which
the corresponding layer of the receiving computer needs in order to process the
incoming packet.
The purpose of the layering structure is to enable
vendors to adapt to specific hardware and software configurations without
recreating the entire stack.
Protocols describe the way in which network data is
encapsulated in packets on the source end, sent via the network to a
destination, and then reconstructed at the destination into the appropriate
file, instruction, or request.
Breaking network data into packet-sized chunks
provides smoother throughput because the small packets don't tie up the
transmission medium as a larger unit of data might. Also, packets simplify the task of error detection and
correction. Each packet is checked
separately for errors, and if an error is discovered, only that packet (instead
of a whole file) must he retransmitted.
The exact composition of a network packet depends on
the protocols you're using. In general,
network packets contain the following:
·
Header. The header signifies the start of the packet and contains a bundle of
important parameters, such as the source and destination address and
time/synchronization information.
·
Data. This
portion of the packet contains the original data being transmitted.
·
Trailer. The trailer marks the end of the packet and typically contains
error-checking (Cyclical Redundancy Check, or CRC) information.
As
the data passes down through the protocol layers, each layer performs its prescribed
function, such as interfacing with an application, converting the data format,
or adding addressing and error-checking parameters. You will soon learn that actual working protocol stacks don't
always comply exactly with the OSI model.
Some, in fact, predate the OSI model, but the concepts and terminology
of the OSI model are nevertheless useful for describing protocol functions.
When the packet reaches the transmission medium, the
network adapter cards of other computers on the network segment examine the
packet, checking the packet's destination address. If the destination address matches the PC's address, the network
adapter interrupts the processor, and the protocol layers of the destination PC
process the incoming packet.
Protocols
and Protocol Layers
Many of the addressing, error-checking,
retransmission, and acknowledgment services most commonly associated with
networking take place at the Network and Transport OSI layers. Protocol suites are often referred to by the
suite's Transport and Network protocols.
In TCP/IP, for instance, TCP is a Transport layer protocol and IP is a
Network layer protocol. (Note, however,
that TCP/IP predates OSI and diverges from OSI in a number of ways.)
IPX/SPX is another protocol suite known by its Transport
and Network layer protocols, but the order of the protocols is backward from
the way the protocols are listed in TCP/IP IPX is the Network layer protocol;
SPX is the Transport layer protocol.
The lower Data Link and Physical layers provide a
hardware-specific foundation, addressing items such as the network adapter
driver, the media access method, and the transmission medium. Transport and Network layer protocols such
as TCP/IP and IPX/SPX rest on that Physical and Data Link layer foundation, and,
with the help of the NDIS and ODI standards, multiple protocol stacks can
operate simultaneously through a single network adapter.
Upper-level protocols provide compatibility with a
particular networking environment. For
instance, the so-called NetBIOS over TCP/IP stack provides Microsoft clients
with TCP/IP.
This chapter describes the common protocol suites
and many of the important protocols associated with them. In addition to TCP/IP
and IPX/SPX, some of the common Transport and Network layer protocols are the
following:
·
NWLink. Microsoft's version of the IPX/SPX protocol essentially spans
the Transport and Network layers.
·
NetBEUI. Designed for Microsoft
networks, NetBEUI includes functions at the Network and Transport layers. NetBEUI isn't routable and therefore doesn't
make full use of Network layer capabilities.
·
AppleTalk Transaction Protocol (ATP) and Name Binding Protocol (NBP).
ATP and NBP are AppleTalk Transport layer protocols.
·
Datagram Delivery Protocol (DDP).
DDP is the AppleTalk Network layer protocol.
Windows NT Networking
Microsoft describes
the Windows NT networking architecture as shown in the figure above. Note the
importance of NDIS in the Windows NT networking structure. The NDIS interface, NDIS wrapper, and NDIS compatible
drivers enable the TCP/IP, NWLink, NetBEUI, AppleTalk, and DLC protocols to
interact simultaneously with the lower layers.
The Transport Driver Interface (TDI) is an interface
that enables the server, redirector, and file system drivers to remain
independent of the transport protocol.
NWLink is Microsoft's version of IPX/SPX.
Windows NT (like other Microsoft operating systems
such as Windows for Workgroups and Windows 95) services client requests by
using the Server Message Block (SMB) protocol.
SMB is an Application layer protocol.
Three stages must take place before a protocol is
operational:
1. A
model describes the general function of the protocol.
2. The
protocol is defined in complete detail.
3. The
protocol must be realised by software and hardware designers in real products.
Consider the process of designing a building: The
architect first produces sketches that describe the general nature of the
building. Then the architect, possibly
working with a specialist in particular building trades, develops blueprints
that describe every detail of the building.
Finally, an actual building is constructed.
Internet Protocols (TCP/IP)
The Internet protocol suite (also commonly called
the TCP/IP protocol suite) was originally developed by the United States
Department of Defence (DoD) to provide robust service on large internetworks
that incorporate a variety of computer types. In recent years, the Inter net
protocols constitute the most popular network protocols currently in use.
One reason for the popularity of TCP/IP is that no
one vendor owns it, unlike the IPX/SPX, DNA, SNA, AppleTalk protocol suites,
all of which are controlled by specific companies. TCP/IP evolved in response to input from a wide variety of
industry sources. Consequently, TCP/IP
is the most open of the protocol suites and is supported by the widest variety
of vendors. Virtually every brand of
computing equipment now supports TCP/IP.
Much of the popularity of the TCP/IP protocols comes
from their early availability on Unix.
The protocols were built into the Berkeley Standard Distribution (BSD)
Unix implementation. Since then,
TCP/IP has achieved universal acceptance in the Unix community and is a
standard feature on all versions of Unix.
The Internet suite doesn't include protocols for the
Data Link or Physical layers. TCP/IP
was designed to work over established standards such as Ethernet. Over time, TCP/IP has been interfaced to the
majority of Data Link and Physical layer technologies.
The Internet protocols do not map cleanly to the OSI
reference model. The DoD model was,
after all, developed long before the OSI model was defined. The model for the Internet protocol suite
has four layers. From this model, you
can see the approximate relationships of the layers. The DoD model's layers function as follows:
The Network Access layer corresponds to the bottom
two layers of the OSI model. This
correspondence enables the DoD protocols to coexist with existing Data Link and
Physical layer standards.
The Internet layer corresponds roughly to the OSI
Network layer. Protocols at this layer
move data between devices on networks.
The Host-to-Host layer can be compared to the OSI
Transport layer. Host-to-Host protocols
enable peer communication between hosts on the internetwork. (At the time these protocols were designed,
personal computers and workstations didn't exist, and all network computers
were host computers. As a result,
devices on TCP/IP networks are typically referred to as hosts. The concept of a client/server relationship
didn't exist, and all communicating hosts were assumed to be peers.)
The Process/Application layer embraces functions of
the OSI Session, Presentation, and Application layers. Protocols at this layer provide network
services.
One huge advantage of TCP/IP is that TCP/IP is
required for communication over the Internet.
One disadvantage is that the size of the protocol stack makes TCP/IP
difficult to implement on some older machines.
(Present day PC models should have no problem running TCP/IP.) TCP/IP has traditionally been considered
slower than other protocol stacks, but again, the power of the newer machines
overcomes much of this difficulty.
A large number of protocols are associated with
TCP/IP. Several of these are discussed
briefly in the following sections.
Internet Protocol (IP)
The Internet Protocol (IP) is a connectionless
protocol that provides datagram service, and IP packets are most commonly
referred to as IP datagrams. IP is a
packet-switching protocol that performs addressing and route section. An IP header is appended to packets, which
are transmitted as frames by lower level protocols. IP routes packets through
internetworks by utilising dynamic routing tables that are referenced at each
hop. Routing determinations are made by
consulting logical and physical network device information, as provided by the
Address Resolution Protocol (ARP).
IP performs packet disassembly and reassembly as
required by packet size limitations defined for the Data Link and Physical
layers being implemented. IP also
performs error checking on the header data using a checksum, although data from
upper layers is not error-checked.
Internet Control Message Protocol
(ICMP)
The Internet Control Message Protocol (ICMP)
enhances the error control provided by IP.
Connectionless protocols, such as IP, cannot detect internetwork errors,
such as congestion or path fail-ures.
ICMP can detect such errors and notify IP and upper4ayer protocols.
Routing Information Protocol (RIP)
The Routing Information Protocol (RIP) in the TCP/IP
suite is not the same protocol as RIP in the NetWare suite, although the two
serve similar functions. Internet RIP performs
route discovery by using a distance-vector method, calculating the number of
hops that must be crossed to route a packet by a particular path.
Although it works well in localized networks, RIP
presents many weaknesses that limit its utility on wide-area
internetworks. RIP's distance-vector
route discovery method, for example, requires more broadcasts and thus causes
more network traffic than some other methods.
The OSPF protocol, which uses the link-state route discovery method, is
gradually replacing RIP.
Open Shortest Path First (OSPF)
The Open Shortest Path First (OSPF) protocol is a
link-state route-discovery protocol that is designed to overcome the
limitations of RIP. On large internetworks, OSPF can identify the internetwork
topology and improve performance by implementing load balancing and
class-of-service routing.
Transmission Control Protocol (TCP)
The Transmission Control Protocol (TCP) is an
internetwork protocol that corresponds to the OSI Transport layer. TCP provides full-duplex, end-to-end
connections. When the overhead of
end-to-end communication acknowledgment isn't required, the User Datagram
Protocol (UDP) can be substituted for TCP at the Transport (host-to-host)
level. TCP and UDP operate at the same
layer.
TCP corresponds to SPX in the NetWare
environment. TCP maintains a logical
connection between the sending and receiving computer systems. In this way, the integrity of the
transmission is maintained. TCP detects
any problems in the transmission quickly and takes action to correct them. The trade-off is that TCP isn't as fast as
UDP.
TCP also provides message fragmentation and
reassembly and can accept messages of any length from upper-layer
protocols. TCP fragments message
streams into segments that can be handled by IP. When used with IP, TCP adds connection-oriented service and
performs segment synchronization, adding sequence numbers at the byte level.
In addition to message fragmentation, TCP can
maintain multiple conversations with upper-layer protocols and can improve use
of network bandwidth by combining multiple messages into the same segment. Each
virtual-circuit connection is assigned a connection identifier called a port,
which identifies the datagrams associated with that connection.
User Datagram Protocol (UDP)
The User Datagram Protocol (UDP) is a connectionless
Transport (host-to-host) layer protocol.
UDP does not provide message acknowledgments; rather; it simply
transports datagrams.
Like TCP, UDP utilizes port addresses to deliver
datagrams. These port addresses,
however; aren't associated with virtual circuits and merely identify local host
processes. UDP is preferred over TCP
when high performance or low network overhead is more critical than reliable
delivery. Because UDP doesn't need to
establish, maintain, and close connections, or control data flow, it generally
outperforms TCP
UDP is the Transport layer protocol used with the
Simple Network Management Protocol (SNMP), the standard network management
protocol used with TCP/IP networks. UDP
enables SNMP to provide network management with a minimum of network overhead.
Address Resolution Protocol (ARP)
Three types of address information are used on
TCP/IP internetworks:
·
Physical addresses. Used by the
Data Link and Physical layers.
·
IP addresses. Provide logical
network and host IDs. IP addresses
consist of four numbers typically expressed in dotted-decimal form. An example of an IP address is
134.135.100.13.
·
Logical node names. Identify
specific hosts with alphanumeric identifiers, which are easier for users to
recall than the numeric IP addresses.
An example of a logical node name is MYHOST.COM.
Given a logical node name, the Address Resolution
Protocol (ARP) can determine the IP address associated with that name. ARP maintains tables of address resolution
data and can broadcast packets to discover addresses on the internetwork. The IP addresses discovered by ARP can be
provided to Data Link layer protocols.
Domain Name System (DNS)
The Domain Name System (DNS) protocol provides name
and ad-dress resolution as a service to client applications. DNS servers enable humans to use logical
node names to access network resources.
File Transfer Protocol (FTP)
The File Transfer Protocol (FTP) is a protocol for
sharing files between networked hosts.
FTP enables users to log on to remote hosts. Logged-on users can inspect
directories, manipulate files, execute commands, and perform other commands on
the host. FTP also has the capability
of transferring files between dissimilar hosts by supporting a file request
structure that is independent of specific operating systems.
Simple Mail Transfer Protocol
(SMTP)
The Simple Mail Transfer Protocol (SMTP) is a
protocol for routing mail through internetworks. SMTP uses the TCP and IP protocols.
SNMP doesn't provide a mail interface for the
user. Creation, management, and
delivery of messages to end users must be performed by an email
application. (The most popular email
application on the Internet is named Eudora.)
Remote Terminal Emulation (TELNET)
TELNET is a terminal emulation protocol. TELNET enables PCs and workstations to
function as dumb terminals in sessions with hosts on internetworks. TELNET implementations are available for most
end-user platforms, including Unix (of course), DOS, Windows, and Macintosh OS.
Network File System (NFS)
Network File System (NFS), developed by Sun
Microsystems, is a family of file-access protocols that are a considerable advancement
over F'TP and TELNET. Since Sun made the NFS specifications available for
public use, NFS has achieved a high level of popularity.
NFS consists of two protocols:
eXternal Data Representation (XDR). Supports encoding of data in a machine-independent format. C programmers use XDR library routines to
describe data structures that are portable between machine environments.
Remote Procedure Calls (RPC). Function as a service
request redirector that determines whether function calls can be satisfied
locally or must be redirected to a remote host.
Calls to remote hosts are packaged for network
delivery and transmitted to RPC servers, which generally have the capability of
servicing many remote service requests. RPC servers process the service requests
and generate response packets that are returned to the service requester.
NetWare
IPX/SPX
The protocols utilised with NetWare. The NetWare protocols have been designed
with a high degree of modularity. This modularity
makes the NetWare protocols adaptable to different hardware and simplifies the
task of incorporating other protocols into the suite. Windows NT doesn't use the IPX/SPX suite to communicate with
NetWare resources. Microsoft instead
developed a clone of IPX/SPX called IPX/SPX Compatible Transport.
IPX/SPX is generally smaller and faster than
TCP/IP. Like TCP/IP, IPX/SPX is
routable.
The internetwork Packet Exchange Protocol (IPX) is a
Network layer protocol that provides connectionless (datagram) service. (IPX was developed from the XNS protocol
originated by Xerox.) As a Network
layer protocol, IPX is responsible for internetwork routing and maintaining
network logical addresses. Routing uses
the RIP protocol (described later in this section) to make route selections.
IPX relies on hardware physical addresses found at
lower layers to provide network device addressing. IPX also uses sockets, or upper-layer service addresses, to
deliver packets to their ultimate destinations. On the client, IPX support is provided as a component of the
older DOS shell and the current DOS NetWare requester.
The Router Information Protocol (RIP) uses the
distance-vector route discover method to determine hop counts to other
devices. Like IPX, RIP was developed
from a similar protocol in the XNS protocol suite. RIP is implemented as an upper-layer service and is assigned a
socket (service address). RIP is based
directly on IPX and performs Network layer functions.
Sequenced Packet Exchange (SPX) is a Transport layer
protocol that extends IPX to provide connection-oriented service with reliable
delivery. Reliable delivery is ensured
by retransmitting packets in the event of an error. SPX is derived from a similar SPX protocol in the XNS network
protocol suite.
SPX establishes virtual circuits called
connections. The connection ID for each
connection appears in the SPX header. A
given upper layer process can be associated with multiple-connection IDs.
SPX is used in situations where reliable transmission
of data is needed. SPX sequences the packets of data. Missing packets or packets that don't arrive in the order in
which they were sent, are detected immediately. In addition, SPX offers connection multiplexing, which is used in
the printing environment.
Many accounting programs, for example, depend upon
the services of SPX to ensure that data is sent accurately. On the client, SPX support is provided as a
component of the older DOS shell and of the current NetWare requester.
The NetWare Core Protocol (NCP) provides numerous
function calls that support network services, such as file service, printing,
name management, file locking, and synchronisation. NetWare client software interfaces with NCP to access NetWare
services.
NCP is a high-level protocol built into the
NetWare operating system kernel. NCP covers aspects of the Session,
Presentation, and Application layers of the OSI reference model and has its own
miniature language that programmers use when writing applications for the NetWare
environment. The commands that NCP
understands are associated primarily with access to files and directories on a
file server.
NetBEUI
NetBEUI is a transport protocol that serves as an
extension to Microsoft's Network Basic Input/Output System (NetBIOS). Because NetBEUI was developed for an earlier
generation of DOS-based PCs, it is small, easy to implement, and fast - the
fastest transport protocol available with Windows NT. Because it was built for small, isolated LANs, however, NetBEUI
is non-routable, making it somewhat outdated in today's diverse and
interconnected networking environment.
Fortunately, the NDIS standard enables NetBEUI to
coexist with other routable protocols. For instance, you could use NetBEUI for
fast, efficient communications on the LAN segment and use TCP/ IP for
transmissions that require routing.
AppleTalk
AppleTalk is the computing architecture developed by
Apple Computer for the Macintosh family of personal computers. Although
AppleTalk originally supported only Apple's proprietary LocalTalk cabling
system, the suite has been expanded to incorporate both Ethernet and Token Ring
Physical layers.
AppleTalk originally supported networks of limited
scope. The AppleTalk Phase 2
specification issued in 1989, however, extended the scope of AppleTalk to
enterprise networks. The Phase 2
specification also enabled AppleTalk to coexist on networks with other protocol
suites.
The LocalTalk, EtherTalk, and
TokenTalk Link Access Protocols (LLAP, ELAP, and TLAP) integrate AppleTalk
upperlayer protocols with the LocalTalk, Ethernet, and Token Ring environments.
Apple's Datagram Deliver Protocol (DDP) is a Network layer protocol that provides connectionless service between two sockets. A socket is the AppleTalk term for a service address. A combination of a device address, network address, and socket uniquely identifies each process.
DDP performs network routing and consults
routing tables maintained by Routing Table Maintenance Protocol (RTMP) to
determine routing. Packet delivery is
performed by the data link protocol operating on a given destination network.
The AppleTalk Transaction Protocol (ATP) is a connectionless Transport layer protocol. Reliable service is provided through a system of acknowledgments and re transmissions. Retransmissions are initiated automatically if an acknowledgment is not received with-in a specified time interval.
ATP reliability is based on transactions. A
transaction consists of a request followed by a reply. ATP is responsible for segment development
and performs fragmentation and reassembly of packets that exceed the
specifications for lower-layer protocols.
Packets include sequence numbers that enable message reassembly and
retransmission of lost packets. Only damaged or lost packets are retransmitted.
The AppleTalk File Protocol (AFP) provides file
services and is responsible for translating local file service requests into
formats required for network file services.
AFP directly translates command syntax and enables applications to
perform format translations. AFP is
responsible for file system security and verifies and encrypts logon names and
passwords during connection setup.
AppleShare is a client/server system for
Macintosh. AppleShare provides three
primary application services:
·
The AppleShare File Server uses AFP to enable users to store and access
files on the network. It logs in users and associates them with network volumes
and directories.
·
The AppleShare Print Server uses NBP and PAP to support network
printing. NBP provides name and address information that enables PAP to connect
to printers. The Apple-Share Print Server performs print spooling and manages
printing on networked printers.
·
The AppleShare PC enables PCs running MS-DOS to access AppleShare
services by running an AppleShare PC program.
Data Link Control (DLC)
The Data Link Control (DLC) protocol does not
provide a fully functioning protocol stack. (Note that DLC is not continuous
with the upper layers.) In Windows NT
systems, DLC is used primarily to access to Hewlett Packard JetDirect
network-interface printers. DLC also
provides some connectivity with IBM mainframes.