Translation

English
English French Actions
Remote Procedure Calls
In the previous sections, we have described several protocols that enable humans to exchange messages and access to remote documents. This is not the only usage of computer networks and in many situations applications use the network to exchange information with other applications. When an application needs to perform a large computation on a host, it can sometimes be useful to request computations from other hosts. Many distributed systems have been built by distributing applications on different hosts and using `Remote Procedure Calls` as a basic building block.
In traditional programming languages, `procedure calls` allow programmers to better structure their code. Each procedure is identified by a name, a return type and a set of parameters. When a procedure is called, the current flow of program execution is diverted to execute the procedure. This procedure uses the provided parameters to perform its computation and returns one or more values. This programming model was designed with a single host in mind. In a nutshell, most programming languages support it as follows :
The caller places the values of the parameters at a location (register, stack, ...) where the callee can access them
The caller transfers the control of execution to the callee's procedure
The callee accesses the parameters and performs the requested computation
The callee places the return value(s) at a location (register, stack, ...) where the caller can access them
The callee returns the control of execution to the caller's
This model was developed with a single host in mind. How should it be modified if the caller and the callee are different hosts connected through a network ? Since the two hosts can be different, the two main problems are the fact they do not share the same memory and that they do not necessarily use the same representation for numbers, characters, ... Let us examine how the five steps identified above can be supported through a network.
The first problem to be solved is how to transfer the information from the caller to the callee. This problem is not simple and includes two sub-problems. The first sub-problem is the encoding of the information. How to encode the values of the parameters so that they can be transferred correctly through the network ? The second problem is how to reach the callee through the network ? The callee is identified by a procedure name, but to use the transport service, we need to convert this name into an address and a port number.
Encoding data
The encoding problem exists in a wide range of applications. In the previous sections, we have described how character-based encodings are used by email and HTTP. Although standard encoding techniques such as ASN.1 [Dubuisson2000]_ have been defined to cover most application needs, many applications have defined their specific encoding. `Remote Procedure Call` are no exception to this rule. The three most popular encoding methods are probably XDR :rfc:`1832` used by ONC-RPC :rfc:`1831`, XML, used by XML-RPC and JSON :rfc:`4627`.
The eXternal Data Representation (XDR) Standard, defined in :rfc:`1832` is an early specification that describes how information exchanged during Remote Procedure Calls should be encoded before being transmitted through a network. Since the transport service enables transfering a block of bytes (with the connectionless service) or a stream of bytes (by using the connection-oriented service), XDR maps each datatype onto a sequence of bytes. The caller encodes each data in the appropriate sequence and the callee decodes the received information. Here are a few examples extracted from :rfc:`1832` to illustrate how this encoding/decoding can be performed.
For basic data types, :rfc:`1832` simply maps their representation into a sequence of bytes. For example a 32 bits integer is transmitted as follows (with the most significant byte first, which corresponds to big-endian encoding).
XDR also supports 64 bits integers and booleans. The booleans are mapped onto integers (`0` for `false` and `1` for `true`). For the floating point numbers, the encoding defined in the IEEE standard is used.
In this representation, the first bit (`S`) is the sign (`0` represents positive). The next 11 bits represent the exponent of the number (`E`), in base 2, and the remaining 52 bits are the fractional part of the number (`F`). The floating point number that corresponds to this representation is :math:`(-1)^{S} \times 2^{E-1023} \times 1.F`. XDR also allows encoding complex data types. A first example is the string of bytes. A string of bytes is composed of two parts : a length (encoded as an integer) and a sequence of bytes. For performance reasons, the encoding of a string is aligned to 32 bits boundaries. This implies that some padding bytes may be inserted during the encoding operation is the length of the string is not a multiple of 4. The structure of the string is shown below (source :rfc:`1832`).

Loading…

User avatar None

New source string

cnp3-ebook / protocols/rpcFrench

New source string a year ago
Browse all component changes

Glossary

English French
No related strings found in the glossary.

String information

Source string location
../../protocols/rpc.rst:6
String age
a year ago
Source string age
a year ago
Translation file
locale/fr/LC_MESSAGES/protocols/rpc.po, string 1