Chapter 24 - Socket Interface

Introduction API The Socket API Sockets and socket libraries Sockets and UNIX I/O The socket API Summary of socket system calls socket close bind Socket address formats listen accept connect send sendto, sendmsg recv, recvfrom, recvmsg Other procedures Sockets and processes Summary

Introduction
The socket is one form of interface between application programs and protocol software Widely available - program portability Used by both clients and servers Extension to UNIX file I/O paradigm

API
Application interactions with protocol software: Passive listen or active open Protocol to use IP address and port number Interface to protocol is call Application Program Interface (API) Defined by programming/operating system Includes collection of procedures for application program

The Socket API

Protocols do not typically specify API API defined by programming system Allows greatest flexibility - compatibility with different programming systems Socket API is a specific protocol API Originated with Berkeley BSD UNIX Now available on Windows 95 and Windows NT, Solaris, etc. Not defined as TCP/IP standard; de factostandard

Sockets and socket libraries

BSD UNIX includes sockets as system calls Other vendors (mostly UNIX) have followed suit Some systems have different API Adding sockets would require changing OS Added library procedures - socket library - instead Adds layer of software between application and operating system Enhances portability May hide native API altogether

Sockets and UNIX I/O

Developed as extension to UNIX I/O system Uses same file descriptor address space (small integers) Based on open-read-write-close paradigm open - prepare a file for access read/write - access contents of file close - gracefully terminate use of file Open returns a file descriptor, which is used to identify the file to read/write/close

The socket API

Socket programming more complex than file I/O Requires more parameters Addresses Protocol port numbers Type of protocol New semantics Two techniques Add parameters to existing I/O system calls Create new system calls Sockets use a collection of new system calls

Summary of socket system calls

socket - create a new socket close - terminate use of a socket bind - attach a network address to a socket listen - wait for incoming messages accept - begin using incoming connection connect - make connection to remote host send - transmit data through active connection recv - receive data through active connection

Socket
descriptor = socket(protofamily, type, protocol) Returns socket descriptor used in subsequent calls protofamily selects protocol family; e.g.: PF_INET - Internet protocols PF_APPLETALK - AppleTalk protocols type selects type of communication SOCK_DGRAM - connectionless SOCK_STREAM - connection-oriented protocol specifies protocol within protocol family IPPROTO_TCP - selects TCP IPPROTO_UDP - selects UDP

Close
close(descriptor) Terminates use of socket descriptor descriptor contains descriptor of socket to be closed

Bind
bind(socket, localaddr, address) Initially, socket has no addresses attached bind selects either local, remote or both addresses server binds local port number for incoming messages client binds remote address and port number to contact server

Socket address formats

Because sockets can be used for any protocols, address format is generic: struct sockaddr { u_char sa_len; /* total length of address */ u_char sa_family; /* family of the address */ char sa_data[14]; /* address */ } For IP protocols, sa_data hold IP address and port number: struct sockaddr_in { u_char sin_len; /* total length of address */ u_char sin_family; /* family of the address */ u_short sin_port; /* protocol port number */ struct in_addr sin_addr; /* IP address */ char sin_zero[8] /* unused */ } First two fields match generic sockaddr structure Remainder are specific to IP protocols INADDR_ANY interpreted to mean "any" IP address

listen
listen(socket, queuesize) Server uses listen to wait for incoming connections socket identifies socket through which connections will arrive (address) New connection requests may arrive while server processes previous request Operating system can hold requests on queue queuesize sets upper limit on outstanding requests

Accept
accept(socket, caddress, caddresslen) Server uses accept to accept the next connection request accept call blocks until connection request arrives Returns new socket with server's end of new connection Old socket remains unchanged and continues to field incoming requests caddress returns struct sockaddr client address; format depends on address family of socket caddresslen returns length of address

Connect
connect(socket, saddress, saddresslen) Client uses connect to establish connection to server Blocks until connection completed (accepted) socket holds descriptor of socket to use saddress is a struct sockaddr that identifies server saddresslen gives length of saddress Usually used with connection-oriented transport protocol Can be used with connectionless protocol Marks local socket with server address Implicitly identifies server for subsequent messages

Send
send(socket, data, length, flags) Used to send data through a connected socket socket identifies socket data points to data to be sent length gives length of data (in bytes) flags indicate special options

Sendto, Sendmsg
sendto(socket, data, length, flags, destaddress, addresslen) sendmsg(socket, msgstruct, flags) Used for unconnected sockets by explicitly specifying destination sendto adds additional parameters: destaddress - struct sockaddr destination address addresslen - length of destaddress sendmsg combines list of parameters into single structure: struct msgstruct { struct sockaddr *m_addr; /* ptr to destination address */ struct datavec *m_vec; /* pointer to message vector */ int m_dvlength; /* num. of items in vector */ struct access *m_rights; /* ptr to access rights list */ int m_alength; /* num. of items in list */ }

recv
recv(socket, buffer, length, flags) Used to receive incoming data through connected socket socket identifies the socket Data copied into buffer At most length bytes will be recved flags give special options Returns number of bytes actually recved 0 implies connection closed -1 implies error

recvfrom, recvmsg
recvfrom(socket, buffer, length, flags, sndraddress, addresslen) recvmsg(socket, msgstruct, flags) Like recvfrom and recvmsg (in reverse!) Address of source copied into sndraddress Length of address in addresslen recvmsg uses msgstruct for parameters

Other procedures
getpeername - address of other end of connection getsockname - current address bound to socket setsockopt - set socket options

Sockets and processes

Like file descriptors, sockets are inherited by child processes Socket disappears when all processes have closed it Servers use socket inheritance to pass incoming connections to slave server processes

Summary
Socket API is de facto standard Originally developed for BSD UNIX Copied to many other systems Sockets are an extension of the UNIX file I/O system Use same descriptor addresses Can (but typically don't) use same system calls Many specific system calls for sockets

Chapter 26(29) - the Domain Name System

Introduction Structure of DNS names DNS naming structure Geographic structure Domain names within an organization Example DNS hierarchy DNS names and physical location Client-server computing DNS and client-server computing DNS server hierarchy Choosing DNS server architecture Name resolution DNS messages DNS servers Using DNS servers DNS caching Types of DNS entries Abbreviations Summary

Introduction
IP assigns 32-bit addresses to hosts (interfaces) Binary addresses easy for computers to manage All applications use IP addresses through the TCP/IP protocol software Difficult for humans to remember: % telnet 204.82.11.70 The Domain Name System (DNS) provides translation between symbolic names and IP addresses

Structure of DNS names

Each name consists of a sequence of alphanumeric components separated by periods Examples: www.ubalt.edu home.ubalt.edu ubmail.ubalt.edu Names are hierarchical, with most-significant component on the right Left-most component is computer name

DNS naming structure

Top level domains (right-most components; also known as TLDs) defined by global authority com edu gov Commercial organization Educational institution Government organization

mil

Military organization

Organizations apply for names in a top-level domain: ubalt.edu microsoft.com Organizations determine own internal structure home.ubalt.edu server1.scitutor.com

Geographic structure
Top-level domains are US-centric Geographic TLDs used for organizations in other countries:
TLD Country

.uk .fr .ch .in

United Kingdom France Switzerland India

Countries define their own internal hierarchy: ac.uk and .edu.au are used for academic organizations in the United Kingdom and Australia

Domain names within an organization

Organizations can create any internal DNS hierarchy Uniqueness of TLD and organization name guarantee uniqueness of any internal name (much like file names in your directories) All but the left-most component of a domain name is called the domain for that name: Name Domain

www.ubalt.edu ubalt.edu server1.ait.ubalt.edu ait.ubalt.edu mail.scitutor.com scitutor.com Authority for creating new subdomains is delegated to each domain Administrator of ubalt.edu has authority to create ait.ubalt.edu and need not contact any central naming authority

Example DNS hierarchy

Page 443, Figure 29.2

DNS names and physical location

DNS domains are logical concepts and need not correspond to physical location of organizations DNS domain for an organization can span multiple networks Ubalt.edu covers all networks at UB www.mis.ubalt.edu is on the 4th floor of Merrick Business School laptop.mis.ubalt.edu could be connected to a network in California

Client-server computing
Clients and servers communicate in distributed computing Client initiates contact to request some remote computation Server waits for clients and answers requests as received Clients are usually invoked by users as part of an end-user application Servers are usually run on central, shared computers

DNS and client-server computing

DNS names are managed by a hierarchy of DNS servers Hierarchy is related to DNS domain hierarchy Root server at top of tree knows about next level servers Next level servers, in turn, know about lower level servers

DNS server hierarchy

Page 446, Figure 29.3

Choosing DNS server architecture

Small organizations can use a single server Easy to administer Inexpensive Large organizations often use multiple servers Reliability through redundancy Improved response time through load-sharing Delegation of naming authority Locality of reference applies - users will most often look up names of computers within same organization

Name resolution
Resolver software typically available as library procedures Implement DNS application protocol Configured for local servers Example - UNIX gethostbyname Calling program is client Constructs DNS protocol message - a DNS request Sends message to local DNS server DNS server resolves name Constructs DNS protocol message - a DNS reply Sends message to client program and waits for next request

DNS messages
DNS request contains name to be resolved DNS reply contains IP address for the name in the request

DNS servers
Each DNS server is the authoritative server for the names it manages If request contains name managed by receiving server, that server replies directly Otherwise, request must be forwarded to the appropriate authoritative server

Using DNS servers

DNS request is forwarded to root server, which points at next server to use Eventually, authoritative server is located and IP address is returned DNS server hierarchy traversal is called iterative resolution Applications use recursive iteration and ask DNS server to handle traversal

DNS caching
DNS resolution can be very inefficient Every host referenced by name triggers a DNS request Every DNS request for the address of a host in a different organization goes through the root server Servers and hosts use caching to reduce the number of DNS requests Cache is a list of recently resolved names and IP addresses Authoritative server include time-to-live with each reply

Types of DNS entries

DNS can hold several types of records Each record includes Domain name Record type Data value A records map from domain name to IP address Domain name - Scitutor Record type - A Data value - 208.82.56.118 Other types: MX (Mail eXchanger) - maps domain name used as e-mail destination to IP address CNAME - alias from one domain name to another Result - name that works with one application may not work with another!

Abbreviations
May be convenient to use abbreviations for local computers; e.g. ubmail for ubmail.ubalt.edu Abbreviations are handled in the resolver; DNS servers only know full-qualified domain names (FQDNs) Local resolver is configured with list of suffixes to append Suffixes are tried sequentially until match found

Summary
Domain Name System maps from computer names and IP addresses Important to hide 32-bit IP addresses from humans DNS names are hierarchical and allocated locally Replication and caching are important performance enhancements DNS provides several types of records