| |
Controller
Redundancy: Clustered Servers
The
area of server clustering extends the concept of data redundancy
and operational availabilty to a new level and is far beyond the
scope of this guide to discuss in depth. Physically, a cluster is
a collection of two or more independent but identical servers that
are able to access a common body of data storage and provide services
to a common set of clients. The objective of clustering is to build
an environment where all the servers in a cluster operate logically
as a single entity. In the context of today’s technology, this generally
means servers connected to a common shared storage sub-system, shared
I/O buses, and a common network for client access.
As a clustered server operates logically in a single management
domain, in which any server can provide any available service to
any authorised client, using a RAID system as the common shared
storage system makes perfect sense. Clustering is simply the equivalent
server process of redundant
controllers in RAID. The prerequisite
requirement for common data access and a unified security model
makes the stability of the shared storage vital. If the storage
system fails every server in the cluster will be unable to operate
effectively. Clustering servers with a non-redundant storage system
defeats the entire principle of clustering and redundancy.
With the relative limitations of modern OS technology, this generally
means that the servers comprising a cluster are of the same architecture
and run the same version of the same operating system. UNIX systems
are more ideally suited to clustering than other network operating
system such as NT or NetWare. Microsoft do have a version of NT
- Microsoft Cluster Server (MSCS) originally referred to as Wolf
Pack - that supports clustering but is limited in comparison to
Digital's VAX or Hewlett-Packard's HP-UX for example.
Cluster
& RAID Benefits
While implementations differ, clusters implemented in conjunction
with a RAID storage system provide three fundamental benefits:
- They
improve application and data availability dramatically over single
server/storage environments.
- They
enable applications to grow beyond the capacity of a single server
or a single storage device.
- They
simplify the management of large or rapidly growing systems and
allow for dynamic growth of data capacity.
Related
topics:
|
