Understanding Clustering and GFS
In some enterprise infrastructures, high-performance, reliable, scalable servers and shared storage are necessary, with minimal downtime. Although RAID offers redundancy and NFS offers shared storage, they have limitations. For example, NFS transfer and access rates are slower than I/O to local disks and can have even slower rates depending on the number of simultaneous connections.
The Red Hat Cluster Suite offers application failover across multiple servers. Common servers that use clustering include web servers, database servers, and file servers such as GFS, or Global File Systems.
GFS is a scalable shared storage solution with I/O performance comparable to local disk access. It is usually combined with clustering to provide even more reliable storage with failover, redundancy, and simultaneous shared access to a GFS filesystem. When combined with clustering, the GFS filesystem is used on one or more file servers acting as the storage pool accessed by all the cluster nodes via a Storage Area Network (SAN). In addition to its ability to scale to meet the storage needs of hundreds or more servers simultaneously, the size of each GFS filesystem can be expanded while still in use.
The easiest way to start using the Red Hat Cluster Suite and Red Hat GFS is to install the packages from RHN using the Cluster Suite and GFS software channels. Refer to Chapter 3 for details on installing all the packages from a child software channel.
After installing the appropriate RPM packages, set up the cluster using the Cluster Configuration Tool (system-config-cluster) before configuring GFS. The exact configuration of Cluster Suite and GFS depends on a great deal of factors including the needs of your infrastructure, budget allocated to the system group, amount of shared storage needed plus extra for future expansion, and what type of application servers are to be run on the cluster servers. Refer to the Documentation and Knowledgebase sections of redhat.com for detailed instructions.