79
Network Organization Concepts Operating Systems I

Network Organization Concepts Operating Systems I

Embed Size (px)

Citation preview

Page 1: Network Organization Concepts Operating Systems I

Network Organization Concepts

Operating Systems I

Page 2: Network Organization Concepts Operating Systems I

Introduction

In almost every corporate and academic computing environment, “stand-alone” computers are linked together in a network to send and receive data among every user of the system. Examples include e-mail, web servers (in-house or Internet), database applications, and more.

Page 3: Network Organization Concepts Operating Systems I

Introduction - 2

A common goal of networks is the sharing of resources (both hardware and software) while at the same time controlling access to them.

There are two basic kinds of operating system configurations in a networked environment: a network operating system or distributed operating system.

Page 4: Network Organization Concepts Operating Systems I

Network Operating System

Network is built on top of local computer operating system.

Users are conscious of the assortment of computers on the network. Access is gained by logging on to a specific remote machine or by transferring data from the remote machine to their own.

Page 5: Network Organization Concepts Operating Systems I

Distributed Operating System

Users need not be aware of every machine on the network.

Remote resources accessed as if local. Total view across multiple computer systems

for controlling and managing resources. Offers several advantages over traditional

operating system environments.

Page 6: Network Organization Concepts Operating Systems I

Basic Terminology

A network is a collection of loosely coupled processors (Chapter 6) interconnected by a communications network.

In a distributed system each host’s own resources are local and other hosts’ resources are remote.

Page 7: Network Organization Concepts Operating Systems I

Basic Terminology - 2

Processors are referred to as sites, hosts, and nodes. “Site” is usually a particular network

location containing one or more computer systems.

“Host” is a specific computer system at a site whose resources can be used from remote locations.

“Node” is a unique name for a computer.

Page 8: Network Organization Concepts Operating Systems I

Basic Terminology - 3

Typically, a host at one site - called the server - has resources that another machine at the same site or a remote site - the client - wants to use.

A server can be a client and a client can also be a server - depending on the situation. (For example, I might log on to a server ...)

Page 9: Network Organization Concepts Operating Systems I

Network Topologies

Sites can be physically or logically connected to one another in a variety of ways (topologies).

Most common topologies are ring, star, bus, tree, and hybrid.

There are tradeoffs between speed, fault tolerance, cost, and difficulty of making a large number of connections.

Page 10: Network Organization Concepts Operating Systems I

Network Topologies - 2

Three criteria need to be kept in mind when deciding on a network topology: Basic cost: the expense of linking the

various sites in the system. Communications cost: time required to

send a message from one site to another. Reliability: assurance that other sites can

still communicate if one of the links fails.

Page 11: Network Organization Concepts Operating Systems I

Star Topology

Also known as “hub” or “centralized topology” - all data passes through a central controller when going from sender to receiver.

Allows easy routing because the central controller knows the paths to all other sites.

Access to network easily controlled.

Page 12: Network Organization Concepts Operating Systems I

Star Topology - 2

Sites can be given priority. However:

Central site must be extremely reliable. Must be able to handle all network traffic,

no matter how heavy.

Page 13: Network Organization Concepts Operating Systems I

Star Topology

CentralController

Host 1Host 5

Host 4

Host 3

Host 2

Page 14: Network Organization Concepts Operating Systems I

Ring Topology

All sites are connected in a closed loop with the first connected to the last.

Packets (containing source and destination address) are passed from one site to the next in one direction.

Destination copies it to a local buffer. Packet continues to circulate until it

arrives back at source, then is deleted.

Page 15: Network Organization Concepts Operating Systems I

Ring Topology - 2

Variations include double loops or rings bridged together: more flexibility, but at a cost.

Disadvantage: all members must be functional for the network to perform properly. (Rings can be designed to allow failed nodes to be bypassed.)

Page 16: Network Organization Concepts Operating Systems I

Bus Topology

All sites are connected through a single communications line.

Sites are connected to the central cable but do not run through a controller.

Only one site can send messages at a time.

End controllers send traffic back down the line in the opposite direction.

Page 17: Network Organization Concepts Operating Systems I

Tree Topology

A tree is a collection of busses. Tree layout begins at “head end” (where

one or more cables start) - each branch may in turn have other branches.

Message circulates through the line, can be received at any other site, and is absorbed at the end points.

Page 18: Network Organization Concepts Operating Systems I

Tree Topology - 2

One advantage of tree and bus topologies is that message traffic still flows through the network even if one node fails.

See Figure 9.8 on page 214.

Page 19: Network Organization Concepts Operating Systems I

Hybrid Topology

Some combination of the other four topologies.

Objective is to take advantage of the strengths of each topology and combine them to most effectively meet the communication requirements.

See Figures 9.9 and 9.10 on page 215 for some examples.

Page 20: Network Organization Concepts Operating Systems I

Network Types

Networks are generally classified by the geographical areas they encompass.

These distinctions are becoming more blurred as advances occur in telecommunications technology.

Basic types are: local area network (LAN), metropolitan area network (MAN), and wide-area network (WAN).

Page 21: Network Organization Concepts Operating Systems I

Local Area Network

Defines a network configuration found within a single building, warehouse, campus, etc.

Typically owned by a single organization and allows computers to communicate through a common communication line.

Page 22: Network Organization Concepts Operating Systems I

Local Area Network - 2

Communications not necessarily confined within the LAN - a gateway or bridge can provide access to the outside.

A bridge is a device (including software) that connects two or more distinct LANs that use the same protocols - e.g., two Ethernet LANs.

Page 23: Network Organization Concepts Operating Systems I

Local Area Network - 3

A gateway is a more complex device (and software) used to connect two or more LANs that use different protocols (e.g., PC network running token-ring and mainframe SNA network).

Physical medium used to construct the network varies. Coaxial cable and fiber optic cable are common.

Page 24: Network Organization Concepts Operating Systems I

Metropolitan Area Networks

Spans an area larger than a LAN, such as several blocks or an entire city (up to a 100 km radius).

High-speed network typically configured as a logical ring.

Messages may be transmitted in one or both directions.

Page 25: Network Organization Concepts Operating Systems I

Wide Area Networks

Connects networks in different parts of the country or even the world.

Uses communications lines of telecommunications companies (like the phone company).

Broad range of communications media, including satellites, microwaves, etc.

The Internet is the best-known WAN.

Page 26: Network Organization Concepts Operating Systems I

Software Design Issues

How do sites use addresses to locate other sites?

How are messages routed? How sent? How do processes communicate with

one another? How are conflicting demands for

resources resolved?

Page 27: Network Organization Concepts Operating Systems I

Addressing Conventions

Network sites are not necessarily connected directly to each other so there is a need for names, addresses, and routes.

Addressing protocols are closely related to the geographic location of the site and its network topology.

Local name and global name may differ.

Page 28: Network Organization Concepts Operating Systems I

Addressing Conventions - 2

Example of a global name: an e-mail address. [email protected]

DNS (Domain Name Service) Protocol is used to resolve Internet addresses (or, more properly, resolve a name to an address).

Page 29: Network Organization Concepts Operating Systems I

Addressing Conventions - 3

DNS works backwards: finds the “edu” domain list. locates the address for “uc” on that list. queries the UC DNS server for host “email”

The name email.uc.edu resolves to an address (129.137.33.153).

E-mail transmission contains that address. E-mail server resolves the name John.Doe.

Page 30: Network Organization Concepts Operating Systems I

Routing Strategies

Routing allows data to get from one point to another on a network.

Each destination must be uniquely identified.

When destination network is reached, the router makes sure the data gets to the correct machine.

Page 31: Network Organization Concepts Operating Systems I

Routing Strategies - 2

Routing protocols need to consider addressing, address resolution, message format, and error reporting.

Most routing protocols use a network and node number to identify each node.

Router stores internally the addresses of networks directly connected.

Routers communicate with one another.

Page 32: Network Organization Concepts Operating Systems I

Routing Strategies - 3

At regular intervals, routers will broadcast their entire routing table.

Addresses allow communication between networks; they do not allow communication between nodes in the same network.

Address resolution - router relates network address to hardware address.

Page 33: Network Organization Concepts Operating Systems I

Routing Strategies - 4

Routers and routing protocols report error conditions (unable to reach destination, etc.) but do not attempt to correct them. (That’s left up to other protocols at other levels of the network architecture.)

Two of the most widely-used routing protocols are routing information protocol (RIP) and open shortest path first.

Page 34: Network Organization Concepts Operating Systems I

Routing Information Protocol

In RIP, selection of a path to a destination is based on the number of intermediate nodes (hops) between the source and destination.

Path with smallest number of hops is always chosen.

Shortest path is not always the best path (may be slow, for example).

Page 35: Network Organization Concepts Operating Systems I

RIP – Continued

Routing table is updated and reissued every 30 seconds, whether or not it has changed.

This leads to increased network traffic and could negatively affect delivery of messages.

May take time for changes to propagate to the other end of the network.

Page 36: Network Organization Concepts Operating Systems I

Open Shortest Path First

In OSPF, selection of a network path is made only after the state of a network has been determined.

Failed intermediate hop eliminated from consideration until it’s back up.

Routing updates are sent only when changes have occurred.

Memory usage is increased (more info).

Page 37: Network Organization Concepts Operating Systems I

OSPF - continued

Savings in bandwidth consumption are offset by increased CPU usage required to compute the shortest path.

Router creates a topological database (data structure) which is updated when failures are reported.

Router checks its database to determine whether a path is available.

Page 38: Network Organization Concepts Operating Systems I

OSPF - continued

If the path is not available, the router uses Dijkstra’s algorithm to generate a “shortest path tree” around the failed link.

Page 39: Network Organization Concepts Operating Systems I

Connection Models

Data entering the network at one point is routed to its destination by being switched from node to node, either by circuit switching or packet switching.

Circuit switching - dedicated communication path established between hosts before transmission begins. If path fails, so does communication.

Page 40: Network Organization Concepts Operating Systems I

Connection Models - 2

(Circuit switching, continued): after a slight delay to set up the connection, the network is transparent to users and data is sent at a fixed rate of speed.

Packet switching - Store-and-forward technology in which messages are divided into equal-sized packets which are then sent, and reassembled at their destination.

Page 41: Network Organization Concepts Operating Systems I

Packet Switching

Packet switching is an effective technology for long-distance communication because it permits transmission between devices that send or receive at different rates.

However - no guarantee that all packets will follow the same path or that they will arrive in the “correct” order.

Page 42: Network Organization Concepts Operating Systems I

Packet Switching - Continued

A “header” containing information about the packet is attached to it before it is sent.

Provides greater line efficiency because a single node-to-node circuit can be shared by several packets and does not sit idle for long periods of time.

Allows “priority” messages.

Page 43: Network Organization Concepts Operating Systems I

Packet Switching - Continued

Two methods of selecting path: datagrams and virtual circuits.

Datagram approach: destination and sequence number added to the unique message identifier.

Each packet is then handled independently - route selected as each packet is put onto the network.

Page 44: Network Organization Concepts Operating Systems I

Packet Switching - Datagrams

At the destination, packets are reassembled in the correct order into one continuous message.

Receiving node requests retransmission of missing or damaged packets.

Two distinct advantages: less congestion by using lighter-used paths more reliable: allows for alternate paths

Page 45: Network Organization Concepts Operating Systems I

Packet Switching - Continued

Virtual circuit approach: destination and packet sequence numbers are not added because a complete circuit between sender and receiver is established before transmission begins.

May be several virtual circuits between nodes. Faster than datagram method - only one routing

decision made.

Page 46: Network Organization Concepts Operating Systems I

Packet Switching - Continued

Disadvantage of virtual circuits: If a node fails then all virtual circuits using

that node become unavailable. Congestion is more difficult to resolve

when circuit experiences heavy traffic.

Page 47: Network Organization Concepts Operating Systems I

Conflict Resolution

Some method is necessary to control access to the network so that all users have equal and fair access.

Access control techniques: round robin, reservation, contention.

Medium access control protocols: carrier sense multiple access, token passing, distributed queue dual-bus.

Page 48: Network Organization Concepts Operating Systems I

Access Control Techniques

Round robin is much like the round robin processor management (chapter 4).

Each node is allowed to use the medium (network). If it has data to send it is given a certain time to complete sending.

If it finishes sooner or at the end of its time, control goes to the next node.

Page 49: Network Organization Concepts Operating Systems I

Access Control (Continued)

Round robin is efficient when there are a lot of nodes transmitting over a long period of time.

Otherwise, overhead of passing control from node to node can be high.

In that case, other techniques may be preferable.

Page 50: Network Organization Concepts Operating Systems I

Access Control (Continued)

Reservation technique access time to medium divided into slots each node can “reserve” a future time slot

for its use. Good for configuration in which several

terminals are connected to a host through a single I/O port.

Page 51: Network Organization Concepts Operating Systems I

Access Control (Continued)

Contention technique Good for short, intermittent traffic No attempt made to control access to the

medium Works well with light to moderate traffic Breaks down under heavy load Easy to implement

Page 52: Network Organization Concepts Operating Systems I

Medium Access Control

Carrier sense multiple access (CSMA) Easy-to-implement, contention-based

protocol. “Carrier sense” - node listens to (tests)

medium before transmitting any data (prevents collision with another node that’s transmitting).

“Multiple Access” - multiple nodes connected to the same medium.

Page 53: Network Organization Concepts Operating Systems I

CSMA (Continued)

Two or more nodes could conclude that all is quiet at the same time, and both begin to transmit at the same time.

Result is damaged messages and unstable situation until bad messages are dissipated.

Failure to receive acknowledgement from receiver causes retransmission.

Page 54: Network Organization Concepts Operating Systems I

CSMA (Continued)

Original CSMA modified to include collision detection - CSMA/CD.

Ethernet is the most common CSMA/CD protocol.

Reduces collisions - does not eliminate them.

Hosts “jammed” when collision detected, retry after a random interval.

Page 55: Network Organization Concepts Operating Systems I

CSMA (continued)

CSMA/CA (collision avoidance) is another modification of CSMA.

Prevents multiple nodes from colliding during transmission.

Used by Apple’s LocalTalk protocol. Sends out a three-byte packet indicating

it wants to transmit. Other nodes wait until this node is done.

Page 56: Network Organization Concepts Operating Systems I

Token Passing

Electronic signal (“token”) is generated when network is turned on.

Token passed from node to node - only the node which has the token can transmit.

Data is sent, with the token. Receiver copies the data, adds the acknowledgement, and returns the packet to the sending node.

Page 57: Network Organization Concepts Operating Systems I

Token Passing

Requires higher overhead at each node than CSMA/CD, and nodes may have long waits under certain conditions.

Token ring is the most widely-used protocol for ring topology. Token moves in one direction from node to

node. When not carrying a message it’s a “free” token.

Page 58: Network Organization Concepts Operating Systems I

Token Passing

Token ring (continued): Nodes waiting to transmit wait for free

token to come by. Free token becomes “busy” and message

sent right after token. Receiving node copies the message, sets

the “copied” bit, and sends the message on around the ring to the sender.

Sender releases a new free token.

Page 59: Network Organization Concepts Operating Systems I

Distributed Queue Dual Bus

Intended for use with a dual bus where each bus sends data in only one direction and has been standardized by one of the IEEE committees as part of its MAN standards.

Transmission is a series of fixed-size slots marked “free” and sent downstream.

Page 60: Network Organization Concepts Operating Systems I

DQDB (Continued)

Free slots are marked “busy” and written to by nodes waiting to send data.

Nodes read and copy data from the slots, which then move to the end of the bus and dissipate.

Very effective protocol even under heavy loads.

Page 61: Network Organization Concepts Operating Systems I

Transport Protocol Standards

Growth of networks in the 1980s and the need to integrate dissimilar network devices from different vendors led to the push for a single universal network architecture.

There are two competing models: OSI and TCP/IP.

Page 62: Network Organization Concepts Operating Systems I

OSI Reference Model

Created by International Organization for Standardization (ISO). Framework for services provided by a

network. Basis for connecting “open” systems for

distributed applications processing. (“Open” systems conform to the model and can be connected regardless of vendor.)

Page 63: Network Organization Concepts Operating Systems I

OSI Reference Model - 2

Once services were identified, some were collected in logical clusters called layers. Allows changes to one layer and its

protocols (for example, due to advances in hardware or software) without affecting the others.

Seven-layer model is software that handles data transmission from one node to another.

Page 64: Network Organization Concepts Operating Systems I

OSI Reference Model - 3

Layer 1 - The Physical Layer Mechanical, electrical, functional

specifications for connecting device to network.

Primary concern is transmitting bits over communication lines.

Only layer concerned with hardware. All data is passed down to layer 1.

Page 65: Network Organization Concepts Operating Systems I

OSI Reference Model - 4

Layer 2 - The Data Link Layer Software for this layer stored in some sort

of device. Establishes and controls the physical path

of communications before sending data to the physical layer below it.

Physically assembles packets for transmission.

Checks for/resolves transmission errors.

Page 66: Network Organization Concepts Operating Systems I

OSI Reference Model - 5

Layer 3 - The Network Layer Provides services such as addressing and

routing. Accepts packets from layer 4, resizes

them, and routes them to the proper destination.

Database of routing tables (dynamically updated) is stored at this level.

Page 67: Network Organization Concepts Operating Systems I

OSI Reference Model - 6

Layer 4 - The Transport Layer Maintains reliable data transmission

between users. Software at this level contains facilities to

handle user addressing and to ensure that all data packets have been received.

Also has a mechanism to control data flow to prevent overrun of slower computer by faster one.

TCP is a layer-4 protocol.

Page 68: Network Organization Concepts Operating Systems I

OSI Reference Model - 7

Layer 5 - The Session Layer Provides user-oriented connection service

and transfers data over communication lines.

Creates and maintains logical link between end points.

User interface. Data flow control. Very similar functions to transport layer.

Page 69: Network Organization Concepts Operating Systems I

OSI Reference Model - 8

Layer 6 - The Presentation Layer Data manipulation functions: compression,

formatting, encryption. Data conversion, syntax conversion, and

protocol conversion are common tasks performed in layer 6.

IBM mainframe CICS teleprocessing monitor is an example (but it also does much more).

Page 70: Network Organization Concepts Operating Systems I

OSI Reference Model - 9

Layer 7 - The Application Layer Application programs, terminals, and

computers access the network at level 7. Provides interface to users and is

responsible for formatting user data before passing it to lower layers.

File transfer and e-mail are two common applications at this layer.

Page 71: Network Organization Concepts Operating Systems I

TCP/IP Model

Most widely used network protocol today. (If you use the Internet you use TCP/IP.)

It is a file transfer protocol so there is a good chance your data will get to its destination without error.

The TCP/IP model organizes a system with three main components: processes, hosts, and networks.

Page 72: Network Organization Concepts Operating Systems I

TCP/IP Model - 2

Model is divided into four layers instead of seven: Network Access Layer - Equivalent to the

physical data link and part of the network layer of the OSI model.

Protocols at this level provide access to a communication network.

Functions performed here include flow control, error control, and security.

Page 73: Network Organization Concepts Operating Systems I

TCP/IP Model - 3

Layer 2 - Internet Layer Equivalent to portion of OSI network layer

not included in Layer 1 of this model, specifically the mechanism that performs routing functions.

Usually implemented within gateways and hosts.

IP is an example of the standard for this layer.

Page 74: Network Organization Concepts Operating Systems I

TCP/IP Model - 4

Layer 3 - Host-to-Host Layer Equivalent to OSI Model Transport and

Session Layers. Supports mechanisms to transport data

between two processes on different host computers.

Services also include error checking and flow control.

Well-know standard: TCP.

Page 75: Network Organization Concepts Operating Systems I

TCP/IP Model - 5

Layer 4 - Process/Application Layer Equivalent to presentation and application

layers of the OSI Model. Includes protocols for computer-to-

computer resource sharing and terminal-to-computer remote access.

Examples: FTP, SMTP, Telnet.

Page 76: Network Organization Concepts Operating Systems I

About the Final Exam

Open book / open notes Covers the entire quarter Objective questions:

Multiple choice Matching

Short Answer questions Exercises in text are good review

questions.

Page 77: Network Organization Concepts Operating Systems I

Taking Operating Systems II?

Emphasis on Windows platform Consumer systems (XP Home Edition) Business workstations (Windows 2000 /

XP Pro) Server and network environment (Windows

2000) What’s new in Windows .NET server

Lecture and hands-on labs (at UC, I hope)

Page 78: Network Organization Concepts Operating Systems I

Taking Operating Systems II?

Overview of Unix Hope to do a Linux lab – we’ll see

Mainframe (OS/390) Enterprise computing Also:

No new text required! Smaller class size!

Page 79: Network Organization Concepts Operating Systems I

A Final Word …

Bring a self-addressed, stamped envelope to the exam if you want your papers back.

Papers are kept for one year, records kept indefinitely.

You can also get papers back if you are in the IT 454 class next quarter.

Please do not e-mail or call me about grades.