Red Hat CLUSTER SUITE FOR ENTERPRISE LINUX 5.1 Uživatelský manuál stáhnout pdf (Strana 22)

Introduction to Linux Clustering

7.3.3 CODA

Coda is an interesting filesystem with features for allowing offline data caching for client

computers, as well as server replication.

Unfortunately, Coda has only really been deployed in research situations and is therefore not

suitable for running in a production environment, but is worth a mention here.

7.4 Replicated Storage

Some distributed and some centralised storage systems have inbuilt methods for data

mirroring (eg: two SANs with hardware mirroring enabled). However, there are also

software solutions that run at the block level and which can mirror any filesystem on top of

them, the most popular one being Distributed Replicated Block Device (DRBD)

7.4.1 DISTRIBUTED REPLICATED BLOCK DEVICE (DRBD)

DRBD is a commonly used to provided block-level disk replication on two-node clusters, by

mirroring the disks between the servers ensuring they have the identical data on them.

Unfortunately, DRBD suffers from the limitation of

only supporting up to two-nodes, although there is a

commercial closed-source version released by the

developers that allows the addition of a third node.

This makes DRBD very useful for creating HA

two-node servers, but not useful for creating a large

shared storage area for large multi-node clusters.

DRBD is an ideal solution for a two-node cluster

that is geographically separated – such as mail or

web servers.

host1

host2

host3

host4

Lustre Storage Pool

host1

host2

sda2

Block level replication

1 2 ... 17 18 19 20 21 22 23 24 25 26 27 ... 32 33

Žádné komentáře

Red Hat CLUSTER SUITE FOR ENTERPRISE LINUX 5.1 Uživatelský manuál Strana 22