Figure 1: Client and server on the same Ethernet communicating using TCP/IP.
How do we build Internet applications? In this lecture, we will discuss the socket API and support for TCP and UDP communications between end hosts. Socket programing is the key API for programming distributed applications on the Internet.
BTW, Kurose/Ross only cover Java socket programming and not C socket programming discussed below.
We plan to learn the following from these lectures:
Program A program is an executable file residing on a disk in a directory. A program is read into memory and is executed by the kernel as ad result of an exec() function. The exec() has six variants, but we only consider the simplest one (exec()) in this course.
Process An executing instance of a program is called a process. Sometimes, task is used instead of process with the same meaning. UNIX guarantees that every process has a unique identifier called the process ID. The process ID is always a non-negative integer.
File descriptors File descriptors are normally small non-negative integers that the kernel uses to identify the files being accessed by a particular process. Whenever it opens an existing file or creates a new file, the kernel returns a file descriptor that is used to read or write the file. As we will see in this course, sockets are based on a very similar mechanism (socket descriptors).
The client-server model is one of the most used communication paradigms in networked systems. Clients normally communicates with one server at a time. From a server’s perspective, at any point in time, it is not unusual for a server to be communicating with multiple clients. Client need to know of the existence of and the address of the server, but the server does not need to know the address of (or even the existence of) the client prior to the connection being established
Client and servers communicate by means of multiple layers of network protocols. In this course we will focus on the TCP/IP protocol suite.
The scenario of the client and the server on the same local network (usually called LAN, Local Area Network) is shown in Figure 1
The client and the server may be in different LANs, with both LANs connected to a Wide Area Network (WAN) by means of routers. The largest WAN is the Internet, but companies may have their own WANs. This scenario is depicted in Figure 2.
The flow of information between the client and the server goes down the protocol stack on one side, then across the network and then up the protocol stack on the other side.
UDP is a simple transport-layer protocol. The application writes a message to a UDP socket, which is then encapsulated in a UDP datagram, which is further encapsulated in an IP datagram, which is sent to the destination.
There is no guarantee that a UDP will reach the destination, that the order of the datagrams will be preserved across the network or that datagrams arrive only once.
The problem of UDP is its lack of reliability: if a datagram reaches its final destination but the checksum detects an error, or if the datagram is dropped in the network, it is not automatically retransmitted.
Each UDP datagram is characterized by a length. The length of a datagram is passed to the receiving application along with the data.
No connection is established between the client and the server and, for this reason, we say that UDP provides a connection-less service.
It is described in RFC 768.
TCP provides a connection oriented service, since it is based on connections between clients and servers.
TCP provides reliability. When a TCP client send data to the server, it requires an acknowledgement in return. If an acknowledgement is not received, TCP automatically retransmit the data and waits for a longer period of time.
We have mentioned that UDP datagrams are characterized by a length. TCP is instead a byte-stream protocol, without any boundaries at all.
TCP is described in RFC 793, RFC 1323, RFC 2581 and RFC 3390.
Socket addresses
IPv4 socket address structure is named sockaddr_in and is defined by including the <netinet/in.h> header.
The POSIX definition is the following:
The uint8_t datatype is unsigned 8-bit integer.
Generic Socket Address Structure
A socket address structure is always passed by reference as an argument to any socket functions. But any socket function that takes one of these pointers as an argument must deal with socket address structures from any of the supported protocol families.
A problem arises in declaring the type of pointer that is passed. With ANSI C, the solution is to use void * (the generic pointer type). But the socket functions predate the definition of ANSI C and the solution chosen was to define a generic socket address as follows:
There are two ways to store two bytes in memory: with the lower-order byte at the starting address (little-endian byte order) or with the high-order byte at the starting address (big-endian byte order). We call them collectively host byte order. For example, an Intel processor stores the 32-bit integer as four consecutives bytes in memory in the order 1-2-3-4, where 1 is the most significant byte. IBM PowerPC processors would store the integer in the byte order 4-3-2-1.
Networking protocols such as TCP are based on a specific network byte order. The Internet protocols use big-endian byte ordering.
The htons(), htonl(), ntohs(), and ntohl() Functions
The follwowing functions are used for the conversion:
The first two return the value in network byte order (16 and 32 bit, respectively). The latter return the value in host byte order (16 and 32 bit, respectively).
The sequence of function calls for the client and a server participating in a TCP connection is presented in Figure 3.
As shown in the figure, the steps for establishing a TCP socket on the client side are the following:
The steps involved in establishing a TCP socket on the server side are as follows:
The socket() Function
The first step is to call the socket function, specifying the type of communication protocol (TCP based on IPv4, TCP based on IPv6, UDP).
The function is defined as follows:
where family specifies the protocol family (AF_INET for the IPv4 protocols), type is a constant described the type of socket (SOCK_STREAM for stream sockets and SOCK_DGRAM for datagram sockets.
The function returns a non-negative integer number, similar to a file descriptor, that we define socket descriptor or -1 on error.
The connect() Function
The connect() function is used by a TCP client to establish a connection with a TCP server/
The function is defined as follows:
where sockfd is the socket descriptor returned by the socket function.
The function returns 0 if the it succeeds in establishing a connection (i.e., successful TCP three-way handshake, -1 otherwise.
The client does not have to call bind() in Section before calling this function: the kernel will choose both an ephemeral port and the source IP if necessary.
The bind() assigns a local protocol address to a socket. With the Internet protocols, the address is the combination of an IPv4 or IPv6 address (32-bit or 128-bit) address along with a 16 bit TCP port number.
The function is defined as follows:
where sockfd is the socket descriptor, servaddr is a pointer to a protocol-specific address and addrlen is the size of the address structure.
bind() returns 0 if it succeeds, -1 on error.
This use of the generic socket address sockaddr requires that any calls to these functions must cast the pointer to the protocol-specific address structure. For example for and IPv4 socket structure:
A process can bind a specific IP address to its socket: for a TCP client, this assigns the source IP address that will be used for IP datagrams sent on the sockets. For a TCP server, this restricts the socket to receive incoming client connections destined only to that IP address.
Normally, a TCP client does not bind an IP address to its socket. The kernel chooses the source IP socket is connected, based on the outgoing interface that is used. If a TCP server does not bind an IP address to its socket, the kernel uses the destination IP address of the incoming packets as the server’s source address.
bind() allows to specify the IP address, the port, both or neither.
The table below summarizes the combinations for IPv4.
| IP Address | IP Port | Result |
| INADDR_ANY | 0 | Kernel chooses IP address and port |
| INADDR_ANY | non zero | Kernel chooses IP address, process specifies port |
| Local IP address | 0 | Process specifies IP address, kernel chooses port |
| Local IP address | non zero | Process specifies IP address and port |
Note, the local host address is 127.0.0.1; for example, if you wanted to run your echoServer (see later) on your local machine the your client would connect to 127.0.0.1 with the suitable port.
The listen() Function
The listen() function converts an unconnected socket into a passive socket, indicating that the kernel should accept incoming connection requests directed to this socket. It is defined as follows:
where sockfd is the socket descriptor and backlog is the maximum number of connections the kernel should queue for this socket. The backlog argument provides an hint to the system of the number of outstanding connect requests that it should enqueue on behalf of the process. Once the queue is full, the system will reject additional connection requests. The backlog value must be chosen based on the expected load of the server.
The function listen() return 0 if it succeeds, -1 on error.
The accept() Function
The accept() is used to retrieve a connect request and convert that into a request. It is defined as follows:
where sockfd is a new file descriptor that is connected to the client that called the connect(). The cliaddr and addrlen arguments are used to return the protocol address of the client. The new socket descriptor has the same socket type and address family of the original socket. The original socket passed to accept() is not associated with the connection, but instead remains available to receive additional connect requests. The kernel creates one connected socket for each client connection that is accepted.
If we don’t care about the client’s identity, we can set the cliaddr and addrlen to NULL. Otherwise, before calling the accept function, the cliaddr parameter has to be set to a buffer large enough to hold the address and set the interger pointed by addrlen to the size of the buffer.
The send() Function
Since a socket endpoint is represented as a file descriptor, we can use read and write to communicate with a socket as long as it is connected. However, if we want to specify options we need another set of functions.
For example, send() is similar to write() but allows to specify some options. send() is defined as follows:
where buf and nbytes have the same meaning as they have with write. The additional argument flags is used to specify how we want the data to be transmitted. We will not consider the possible options in this course. We will assume it equal to 0.
The function returns the number of bytes if it succeeds, -1 on error.
The receive() Function
The recv() function is similar to read(), but allows to specify some options to control how the data are received. We will not consider the possible options in this course. We will assume it equal to 0.
receive is defined as follows:
The function returns the length of the message in bytes, 0 if no messages are available and peer had done an orderly shutdown, or -1 on error.
The close() Function
The normal close() function is used to close a socket and terminate a TCP socket. It returns 0 if it succeeds, -1 on error. It is defined as follows:
There are some fundamental differences between TCP and UDP sockets. UDP is a connection-less, unreliable, datagram protocol (TCP is instead connection-oriented, reliable and stream based). There are some instances when it makes to use UDP instead of TCP. Some popular applications built around UDP are DNS, NFS, SNMP and for example, some Skype services and streaming media.
Figure 4 shows the the interaction between a UDP client and server. First of all, the client does not establish a connection with the server. Instead, the client just sends a datagram to the server using the sendto function which requires the address of the destination as a parameter. Similarly, the server does not accept a connection from a client. Instead, the server just calls the recvfrom function, which waits until data arrives from some client. recvfrom returns the IP address of the client, along with the datagram, so the server can send a response to the client.
As shown in the Figure, the steps of establishing a UDP socket communication on the client side are as follows:
The steps of establishing a UDP socket communication on the server side are as follows:
In this section, we will describe the two new functions recvfrom() and sendto().
The recvfrom() Function
This function is similar to the read() function, but three additional arguments are required. The recvfrom() function is defined as follows:
The first three arguments sockfd, buff, and nbytes, are identical to the first three arguments of read and write. sockfd is the socket descriptor, buff is the pointer to read into, and nbytes is number of bytes to read. In our examples we will set all the values of the flags argument to 0. The recvfrom function fills in the socket address structure pointed to by from with the protocol address of who sent the datagram. The number of bytes stored in the socket address structure is returned in the integer pointed by addrlen.
The function returns the number of bytes read if it succeeds, -1 on error.
The sendto() Function
This function is similar to the send() function, but three additional arguments are required. The sendto() function is defined as follows:
The first three arguments sockfd, buff, and nbytes, are identical to the first three arguments of recv. sockfd is the socket descriptor, buff is the pointer to write from, and nbytes is number of bytes to write. In our examples we will set all the values of the flags argument to 0. The to argument is a socket address structure containing the protocol address (e.g., IP address and port number) of where the data is sent. addlen specified the size of this socket.
The function returns the number of bytes written if it succeeds, -1 on error.
There are two main classes of servers, iterative and concurrent. An iterative server iterates through each client, handling it one at a time. A concurrent server handles multiple clients at the same time. The simplest technique for a concurrent server is to call the fork function, creating one child process for each client. An alternative technique is to use threads instead (i.e., light-weight processes). We do not consider this kind of servers in this course.
The fork() function
The fork() function is the only way in Unix to create a new process. It is defined as follows:
The function returns 0 if in child and the process ID of the child in parent; otherwise, -1 on error.
In fact, the function fork() is called once but returns twice. It returns once in the calling process (called the parent) with the process ID of the newly created process (its child). It also returns in the child, with a return value of 0. The return value tells whether the current process is the parent or the child.
Example
A typical concurrent server has the following structure:
When a connection is established, accept returns, the server calls fork, and the child process services the client (on the connected socket connfd). The parent process waits for another connection (on the listening socket listenfd. The parent closes the connected socket since the child handles the new client. The interactions among client and server are presented in Figure 5.
We now present a complete example of the implementation of a TCP based echo server to summarize the concepts presented above. We present an iterative and a concurrent implementation of the server.
echoClient.c source: echoClient.c
echoServer.c source: echoServer.c
conEchoServer.c source: conEchoServer.c
