Advanced Network Programming Chapter 1 Introduction to Transport Layer and TCP

Advanced Network Advanced Network ProgrammingProgramming

Chapter 1Chapter 1

Introduction

to

Transport Layer and TCP

October, 2001 A. Özgit - Advanced Network Programming 2

Networking (OSI) Reference Networking (OSI) Reference ModelModel

OSI 7-layer Reference Model Application (WEB, NFS, FTP, Telnet, etc.) Presentation (Conversion, Compression,

Cryptography) Session (Synchronization) Transport (End-to-end Messages) Network (Packet Routing) Data Link (Station-to-station Frames) Physical (Bit Transmission)


NetworkingNetworking

Distinction between service and protocol is important! This will be discussed later.

Some widely known transport protocols UDP, TP0, TP4, SNA-APPN, DECnet-NSP, ATM,

XTP, T/TCP, RTP, VMTP, NETBLT


Transport Layer (Layer-4)Transport Layer (Layer-4)

Lowest layer that operates on an end-to-end basis.

Lies at the boundary between hosts and an internetwork of routers, bridges, and communication links.

A good transport layer service Allows applications to use a standard set of

primitives. Run on variety of networks – w/o worrying about

different network interfaces and reliabilities.

Isolates applications from the technology.



Layer-4 provides interprocess communication between two processes that most often are running on different hosts.

TCP and its companion UDP (User Datagram Protocol) are the most widely used protocols.

Other are from IBM’s SNA, and Digital’s (Compaq) DECnet. Connection to proprietary protocol suites.

Ongoing research tcp-impl WG of IETF (www.ietf.org) end2end WG of IRTF (www.irtf.org)



Basic Issues Addressing Connection-oriented (CO) vs. Connectionless

(CL) Reliability

Loss Duplicate Ordering Integrity

Blocking vs. Non-Blocking Multicast, Unicast Priority Security Status Reporting


Role of TCPRole of TCP

The Web An example of client/server application

Web browser (client) Runs on “local” machine Communicates with a server on some “remote”

machine Uses an application layer protocol called the

HTTP (Hypertext Transfer Protocol). HTTP is a simple request/response protocol. We will use HTTP 0.9 (the simplest) in

examples.


Role of TCPRole of TCP

Web browser (client) Access TCP’s service thru function calls that

comprise that Transport Layer’s Application ++++-Programming Interface (API).

API provides (at a minimum) functions to send and receive

messages e.g. Berkeley Sockets – read(), write()

Connection setup and close for CO protocols e.g. connect(), close()


TerminologyTerminology

Simplified Communication Model (OSI) – Figure-1 User Sender / User Receiver at the top Application Entities use the services of the

transport layer Peer Entities exchange Protocol Data Units (PDUs)

APDU The request “get /index.html”

sent from client (application entity) to the server (its peer application entity).

Bi-directional Protocol Both sides can send and receive data

simultaneously.



Transport Entity Hardware and/or software within a given host

that implements a particular transport service and protocol.

User Sender submits a chunk of user data (Transport Service

Data Unit – TSDU; informally a message) to the transport sender.

Transport Sender transmits or sends this data to the transport

receiver over a network which may provide different levels of reliability.



Transport Receiver receives the data that arrives from the network

and delivers it to the user receiver.

TPDUs may flow in both directions even when user data flows only from sender to receiver Control TPDUs

Separate and/or Piggybacked


Terminology Terminology

What happens to the request APDU? APDU becoming a single TSDU, being

encapsulated in a single TPDU, which in turn becomes a single NSDU, which is encapsulated in a single NPDU (Figure-2)

TCP’s TPDU TCP segment Packet IP datagram (NPDU) or TCP segment

(informally)

IP’s PDU Datagram Datagram IP’s NPDU or UDP’s TPDU

(informally)


Example TCP Connection (1)Example TCP Connection (1)

Enter “http://ozgit.nom.tr/index.html” from web client. http indicates application layer protocol to be

used. TCP port number 80 (implicitly) to be used. “ozgit.nom.tr” is the host name (mapped to

an IP number –144.122.71.91- by DNS) Transport Service Access Point (TSAP)

TSAP IP Address + TCP Port Number One end point of a communication channel between a

process on a local m/c and a process on a remote m/c. “index.html” is the file being requested.

http request (APDU) “GET /index.html”

http://ozgit.nom.tr/index.html



Connection request to the transport entity at (144.122.71.91, 80). By calling connect() Local TCP initiates a 3-way handshake with the

remote server. TPDUs are exchanged between TCP entities to

ensure reliable connection establishment and to establish initial sequence numbers.

If 3-way handshake fails, TCP notifies the application.

Otherwise success code is returned -confirmation.

OSI Model: Request – Indication Response - Confirmation



Web client submits a request to send data (APDU – “GET /index.html”)

Local TCP sends this data most likely in a single TPDU. TCP Segment TSDU + Transport Layer

Header



Remote TCP receives the TPDU, the data (APDU – “GET /index.html”) is buffered. Delivered when Web server does a read() This delivery is known as a data indication in

OSI terminology.

Remote TCP also sends back an acknowledgement (ACK) -control TPDU- to the local TCP



The Web server responds with contents of “index.html”. File may be too large to be efficiently submitted

to TCP in one write() call –i.e., one TSDU. Web Server divides APDU into multiple write() calls

–i.e., multiple TSDUs. Remote TCP then sends these TSDUs to local the TCP

in multiple TPDUs.

TCP treats the data as a byte stream and segments it as necessary –i.e., does not care about TSDU boundaries. Boundaries between APDUs, submitted TSDUs,

TPDUs, and delivered TSDUs may all be different.



TCP must detect and recover from network errors. As the remote TCP send the TPDUs, it includes a

sequence number in each TPDU. It also copies each TPDU into a buffer, and sets a

timer. Retransmits the TPDU if timer expires before getting an

ACK. Retransmission is done in a new TPDU. Individual byte-stream sequence numbers are used.

TPDUs retransmitted may or may not correspond exactly to the original TPDUs.

Remote TCP also places a checksum is the TPDU header to detect bit errors.



As TPDUs are received by the local TCP TPDUs with checksum errors are discarded. It ensures that no pieces of the byte-stream are

missing Out-of-order arrivals are reordered. It responds to the remote TCP with ACK TPDUs. Duplicates are discarded (e.g., as a

consequence of lost ACK TPDUs).

Pieces of byte-stream are buffered in local TCP Web client requests them by doing read() calls. Each read() results in delivery of a TSDU.



TCP connection is bi-directional. Either side may initiate the closing of the

connection In first generation web systems the server

initiates the close by calling close() function (Disconnect Request).

Disconnect is handled with a 4-way handshake procedure.


Transport ServiceTransport Service

A transport service abstracts a set of functions that is provided to a higher layer.

A protocol, refers to the details of how a transport sender and a transport receiver cooperate to provide that service.

Distinction between service and protocol is important (Contribution of OSI Reference Model).


CO-message vs. CO-byte vs. CO-message vs. CO-byte vs. CLCL

Two types of transport services Connection-oriented (CO)

Provides for the establishment, maintenance, and termination of a logical connection between transport users (three distinct phases of operation).

Connection Establishment (T-Connect) Data Transfer (T-Data) Connection Termination (T-Disconnect)

CO service has two variations Message-oriented (TP4) Byte-stream

Connectionless (CL) Provides only one phase of operation: data transfer.


ReliabilityReliability

A service is reliable if and only if it satisfies all of the following: No-loss No-duplicates Ordered Data Integrity


No-loss vs. Uncontrolled-loss No-loss vs. Uncontrolled-loss vs. Controlled-lossvs. Controlled-loss

No-loss (at-least-once delivery) service guarantees either of the two results: The data is delivered to the user receiver, or The user sender is notified that some data may

not have been delivered.

Uncontrolled-loss (best-effort) No assurance Example: UDP

Controlled-loss Loss may occur, but there is control over the

degree of loss. Example: k-XP


No-duplicates vs. Maybe-No-duplicates vs. Maybe-duplicatesduplicates

No-duplicates At-most-once delivery e.g., TCP

Maybe-duplicates Efforts by the protocol may or may not be made

to avoid delivering duplicates. e.g., UDP


Ordered vs. Unordered vs. Ordered vs. Unordered vs. Partially-orderedPartially-ordered

Ordered service Preserves user sender’s submission order of

data. e.g., TCP

Unordered service Does not provide the above guarantee. e.g., UDP

Partially-ordered service Guarantees to deliver pieces of data in one of a

set of permitted orders as predefined by a partial order relation agreed upon by the user sender and user receiver.

e.g., Multimedia comm., distributed databases.


Data-integrity vs. No-data-Data-integrity vs. No-data-integrity vs. Partial-data-integrity vs. Partial-data-

integrityintegrity Data-integrity

Ensures with high probability that all data bits delivered to a user receiver are identical to those originally submitted.

Strength of the error detection method. TCP uses 16-bit checksum.

No-data-integrity Provide no guarantees regarding bit errors.

Partial-data-integrity A controlled amount of bit errors (as a means of

achieving higher throughput). e.g., real-time multimedia application


Remarks on Reliability and CO Remarks on Reliability and CO vs. CLvs. CL

All aspects of reliability (loss, duplicates, order, data-integrity) are orthogonal. Data might get lost while the order is

preserved.

Relationship between a service being CO or CL and whether or not it is reliable. These two services are orthogonal. CO service is assumed to be reliable. Why?



Whereas: TCP service is CO and TCP service is reliable,

Whereas: TP4 service is CO and TP4 service is reliable,

Whereas: X.25 service is CO and X.25 service is reliable

CO service Reliable Service ?



Whereas: UDP service is CL and UDP service is unreliable,

CL service Unreliable Service ?


Blocking vs. Non-blockingBlocking vs. Non-blocking

Blocking service Ensures that the transport layer is not

overwhelmed with incoming data. Provides flow control between user sender and

transport sender.

Non-blocking service Allows the user sender to submit data and

continue processing w/o awaiting the transport sender’s OK.


Multicast vs. UnicastMulticast vs. Unicast

Multicast service Enables a user sender to submit data, a copy of

which will be delivered to one or more user receiver(s).

Unicast service Delivery of data to exactly one user receiver.


Priority vs. No-priorityPriority vs. No-priority

Priority service Enables a user sender to indicate the relative

importance of various messages. May be combined with uncontrolled-loss or

controlled-loss service to drop lower priority data, thereby allowing the delivery of high-priority data with smaller delay and/or higher probability.

No-priority service No differentiation of the importance of the

classes of data.


Security vs. No-securitySecurity vs. No-security

Security service A security service provides one or more security

functions such as authentication, access control, confidentiality, and integrity [ISO-7498-2].

Authentication is the verification of user sender’s and user receiver’s identity.

Access control checks a user’s permission status, allowing the use of different resources.

Confidentiality guarantees that only the intended user receiver(s) can decode and understand the user sender’s data.

Integrity detects any modification, insertion, deletion, or replay of transport sender’s data.

e.g., TP4

No-security service Does not provide any of the above security

functions.


Status-reportingStatus-reporting vs. vs. Non-status-Non-status-reportingreporting

Status-reporting service Allows a user sender to obtain specific

information about the transport entity or its connections.

Non-status reporting service Does not provide any information about the

transport entity and its connections.


QQooS S vs. vs. No-QNo-QooSS

QoS service Allows a user sender to specify the quality of

transmission service desired.

No-QoS service Delivery of data to exactly one user receiver.


QQooS S Parameters (ISO)Parameters (ISO)

Connection Establishment Delay

Connection Establishment Failure Probability

Throughput

Transit Delay

Residual Error Rate

Transfer Failure Probability

Connection Release Delay

Connection Release Failure Probability


QQooS S Parameters (ISO)Parameters (ISO)

Protection

Priority

Resilience

The ATM environment supports only two QoS parameters: (sustained) target, acceptable, and minimum

throughput Transit delay


Transport Protocol FeaturesTransport Protocol Features

CO vs. CL Establishment and maintenance of state

information A record of characteristics and events related to the

communication between the transport sender and receiver.

CO: state information is maintained Three phases:

Connection Establishment Data Transfer Connection Termination

CL: no state information is maintained



Transaction Oriented A single APDU (request) is sent by user sender User receiver responds with a single APDU

(response) Characteristics:

Asymmetrical model Simplex data transfer Short duration Low delay Few data TPDUs Message orientation Need for a no-duplicate service



CO Protocol Features Signaling – exchange of control (state)

information In-band (more suitable for short-lived connections) Out-of-band (desirable for high-speed communication

systems)

Unidirectional vs. Bidirectional



Connection Establishment (See Figure-3) Implicit connect

Connection is established as soon as the first TPDU is sent or received.

2-way-handshake connect CR-TPDU (Connection Request) CC-TPDU (Connection Confirm)

3-way-handshake connect CR-TPDU (Connection Request) CC-TPDU (Connection Confirm) ACK-CC-TPDU (ACK for Connection Confirm)



Connection Termination (See Figure-4) Implicit disconnect

Time-out Abortive disconnect

Close connection abnormally due to an error condition 2-way-handshake disconnect

DR-TPDU (Disconnect Request) DC-TPDU (Disconnect Confirm)

4(3)-way-handshake disconnect Two 2-way-handshakes one for each direction of data

flow 3-way if the first DC-TPDU also functions as a DR-

TPDU for the reverse direction


Error ControlError Control

Guard against loss or damage of user data and control information

For realistic high-speed networks with low error rates, transport layer error control is more efficient than link layer error control.

Two phases: Error detection Error reporting and recovery



Error Detection Identifies lost, misordered, duplicated and

corrupted TPDUs Sequence numbers handles the first three

problems Corrupted data is discovered by means of:

Length fields Error Detecting Codes (EDC)

The header/trailer, the data, or the both Separate EDCs are recommended for multimedia

applications



Error Reporting and Recovery Error reporting is a mechanism where receiver

explicitly informs the sender about errors that have been detected.

Error recovery is a mechanism used by both sender and receiver to recover from errors whether or not they are explicitly reported.

Timers, sequence numbers and acknowledgements are used.



Error Reporting and Recovery A positive ACK (PACK) PAR (Positive ACK with Retransmission) or ARQ

(Automatic Repeat Request) Upon receipt of an ACK, the sender updates its state

information, discards buffered TPDUs that are acknowledged, and retransmits any TPDUs that are not acknowledged.

In case of timeout, it may assume something has gone wrong and retransmits unacknowledged TPDU(s).

No error reporting mechanism



Error Reporting and Recovery A negative ACK (NACK) aka Selective Reject

Explicitly identifies TPDUs that have not been received



Piggybacking Artificially delay returning an ACK hoping the

receiver will soon submit its next message to be sent as a part of the reverse direction data flow.

When this occurs, the ACK is piggyback-ed as header information on the reverse direction data TPDU.



Cumulative vs. Selective Acknowledgement Cumulative PACK

Carries a sequence number indicating that all TPDUs with lower sequence numbers have been received.

A recent cumulative PACK incorporates the information of the previously lost one.

Unnecessary retransmissions of correctly received TPDUs.

Selective PACK Acknowledges exactly one TPDU

Block PACK Variation of selective PACK where blocks of individual

TPDUs are selectively acknowledged.



Retransmission Strategies When the sender does not receive a PACK within a

pre-determined timeout period, or when it receives back-to-back cumulative PACKs that are identical.

Selective Repeat (Conservative) Sender retransmits selectively only TPDUi and wait for a

PACK with sequence number larger than previous PACKs.

Go-Back-N (More Aggressive) Sender retransmits TPDUi and all TPDUs already sent

after TPDUi Decrease channel utilization by potentially retransmitting

correctly-received TPDUs.


Flow ControlFlow Control


Multiplexing/DemultiplexingMultiplexing/Demultiplexing

See Figure 5(a)

Several transport layer connections using a single network layer association.

Efficient use of network layer resources.


TCP TCP (Transmission Control (Transmission Control Protocol)Protocol)

Connection-oriented (CO)

Documents

Advanced Network Programming Chapter 1 Introduction to Transport Layer and TCP