||Donald K. Burleson
Oracle RAC Tips
Cluster File System
A cluster file system provides an alternative to raw devices. A file
system presents a familiar environment to the user for creating and
managing the files required for a RAC database. The cluster file system
provides a view of the database files and makes it easy to expand, add,
and copy existing files. With a shared file system, multiple instances can
share archive logs. This makes media recovery much more convenient because
every node has access to every archived log file as needed. The general
For applications, the CFS
functionality is just like a regular file system. The RDBMS system is
easily integrated with the CFS and requires no special modifications.
Most of the CFS solutions are
based on a master-server design, where one of the nodes acts as a server
and the others act as clients. The master node performs all the file
system metadata manipulations, with other nodes forwarding the tasks to
the server node (e.g., Veritas CFS).
Polyserve Matrix Server is
based on symmetric architecture with a distributed lock manager. This
means that all the nodes in the cluster collectively undertake the
metadata and lock management operations. No bottleneck and no single
point of failure.
When using a CFS, Oracle9i
RAC requires Direct I/O. Direct I/O is generally enabled for a file by
opening the file with the o_direct flag in the open(2) system call..
Read and write requests for the file are then executed to and from disk
storage through direct memory access thus bypassing CFS caching. This
improves the I/O performance for database applications that perform
their own caching and file region synchronization. This is, however, a
compile-time requirement. To that end, PolyServe Matrix Server features
a mount-time Direct I/O that enables any application that is capable of
using raw partitions to benefit from Direct I/O without recompiling.
Without the Direct I/O
feature (or the ODM interface), maintenance and synchronization of the
local file system cache becomes complicated. A typical CFS provides an
internal locking mechanism to control the concurrent use and protect the
integrity of the file system (metadata). A locking mechanism, usually
called the lock manager, is used for metadata integrity and cache
coherency across the multiple nodes. The lock manager provides a way to
ensure all nodes have a consistent view of the file system. The lock
manager can be of distributed or master-slave architecture.
One of the advantages that
accrue from using the cluster file system is the combination of OMF with
the ODM interface. The cluster file systems that are compliant with the
ODM interface are able to provide much better I/O performance, with
relative ease of administration. Oracle managed files are able to create
and delete files directly at the operating system level. We will discuss
OMF and ODM again later.
For more information,
see the book
Grid and Real Application Clusters 30% off if you
buy it directly from Rampant TechPress .
Written by top Oracle
experts, this RAC book has a complete online code depot with
ready to use RAC scripts.