|
Distributed operating systemsAfter creating a file, for example the file, /machine1/usr/ast/xyz, the programmer, process, or library can use the cp command (or equivalent) to make copies in /machine2/usr/ast/xyz and /machine3/usr/ast/xyz. Programs can be written to accept strings like /usr/ast/xyz as arguments, and successively try to open the copies until one succeeds. While this scheme can be made to work, it is a lot of trouble. For this reason, a distributed system should do better. In Fig. 5-12(b) we see an alternative approach, lazy replication. Here, only one copy of each file is created, on some server. Later, the server itself makes replicas on other servers automatically, without the programmer's knowledge. The system must be smart enough to be able to retrieve any of these copies if need be. When making copies in the background like this, it is important to pay attention to the possibility that the file might change before the copies can be made. Our final method is to use group communication, as shown in Fig. 5-13(c). In this scheme, all write system calls are simultaneously transmitted to all the servers, so extra copies are made at the same time the original is made ...» | Код для вставки книги в блог HTML
phpBB
текст
|
|