Red Hat CLUSTER SUITE FOR ENTERPRISE LINUX 5.1 Instrukcja Użytkownika Strona 21

  • Pobierz
  • Dodaj do moich podręczników
  • Drukuj
  • Strona
    / 33
  • Spis treści
  • BOOKMARKI
  • Oceniono. / 5. Na podstawie oceny klientów
Przeglądanie stron 20
Introduction to Linux Clustering
7.3 Distributed Storage
Distributed storage takes another approach to the storage mechanism, and instead of having a
central location of data, the data is spread across all the nodes in the cluster, often including
some form of redundancy in order to be able to cope with the failure of a node.
Unfortunately, this redundancy comes at a cost – distributed storage solutions are complex,
and have to be able to handle issues such as the failure of nodes, delays in the network
linking the machines and locking issues.
Note: You may have come across DRBD, which is a two-node block replication solution.
This is covered in the next section of this document, under “Replicated Storage”.
7.3.1 AFS – ANDREW FILE SYSTEM
AFS is a distributed filesystem which caches data locally on machines. Currently there are
two different implementations of AFS for Linux:
OpenAFS (IBM Public License)
AFS implementation in vanilla kernel (under development)
The caching provides increased speed and limited off-line access in the event of network
failures, but the servers do not replicate themselves (although that could be achieved with
DRBD).
However, due to the types of file locking used, it is not suitable for large shared databases,
and can not handle a single file being updated by multiple clients.
AFS was designed to run services such as mail servers using maildir where each email is
stored as an individual file.
7.3.2 LUSTRE
Lustre is a distributed filesystem suitable at creating massive (many thousands of nodes)
distributed filesystems.
Lustre is quite a complex technology to setup and unfortunately does not provide it's own
data replication system. If data replication is required, then another technology like DRBD is
needed to perform the replication between individual nodes, which does limit the scalability.
© Copyright 2008 Jethro Carr Page 21/33
Przeglądanie stron 20
1 2 ... 16 17 18 19 20 21 22 23 24 25 26 ... 32 33

Komentarze do niniejszej Instrukcji

Brak uwag