Red Hat CLUSTER SUITE FOR ENTERPRISE LINUX 5.1 Instrukcja Użytkownika Pobierz pdf (Strona 16)

Introduction to Linux Clustering

7 Storage Management

Storage management may appear to be a separate topic, but it is in fact a very important part of a

cluster's design.

For a cluster, it is very important that data remains intact and accessible by all the nodes. One major

topic is the use of a cluster-capable filesystem such as GFS.

It is also important to choose the correct storage media for the cluster, taking future growth into

consideration. Is performance or reliability more important? Are all the nodes in the same premises,

or do you require a distributed storage solution that will work across the internet? Does the data

need to be replicated in real time between the nodes?

There is a vararity of solutions available (such as SANs), however all solutions fit into one of the

following three categories:

● Centralised storage.

● Distributed storage.

● Replicated storage (this can sometimes be a feature of either of the two categories above)

7.1 Centralised storage

Centralised storage involves having one or more devices providing storage to all the other

nodes. A typical example is a single array of disks such as SAN or NAS, which all the nodes

connect to for storage.

Centralised storage solutions are often found in enterprise server installations, with many

medium to large organisations using something like a SAN for their data storage needs.

Centralised storage is popular for a number of reasons:

● More cost effective to purchase a single array of disks than purchasing disks for each

server.

● Central location allows for easier backups and mirroring.

● Easy to configure, easy to maintain – if you need to add more storage, there's only

one device to upgrade.

However, there is a common problem with centralised storage - Often there will be just a

single device providing the storage. (often due to the cost of purchasing redundant hardware

being too high, devices such as SANs are not cheap).

This introduces a single point of failure – if a hardware fault occurs in the device, it could

cripple the entire cluster, since all nodes rely on it. To prevent this, you either need to be

prepared for the possibility (and cost due to downtime) of a device failure or invest in

redundant hardware.

1 2 ... 11 12 13 14 15 16 17 18 19 20 21 ... 32 33

Brak uwag

Red Hat CLUSTER SUITE FOR ENTERPRISE LINUX 5.1 Instrukcja Użytkownika Strona 16