Oracle Real Application Clusters 10g

What Is Oracle Real Application Clusters 10G?unused IP address on the same subnet as LAN. This
Oracle Real Application Clusters is an option of Oracleaddress is used by application to connect to the RAC
Database that was first introduced with Oracledatabase. If a node fails, the virtual IP is failed over
9i.Oracle Real Application Clusters is now provento another node in the cluster to provide an
technology used by thousands of customers in everyimmediate note down response to connection
industry in every type of application. Oracle RACrequests.
provides option for scaling applications beyond theBenefits of Real Applications Clusters
capabilities of a single server. This allows customersScalability
to take advantage of lower cost commodityScalability is the ability to add additional nodes to Real
hardware to reduce their total cost of ownership andApplication Clusters and achieve markedly improved
provide a scalable computing environment thatperformance. Real Application Clusters can take
supports their application workload. Oracle RACadvantage of additional equipment and harness the
provides the ability to remove the server as a singleprocessing power of multiple systems.
point of failure in any database applicationAll servers in the cluster must run the same operating
environment.system and same version of oracle but they do not
Real Application Clusters Architecturehave to be exactly the same capacity.
A RAC database is a clustered database. A cluster isHigh Availability
a group of independent servers that cooperate as aThe term high availability refers to systems with
single system .Clusters provide improved faultredundant components that provide consistent,
resilience and modular incremental system growthuninterrupted service, even in the event of hardware
over single symmetric multiprocess(SMP) systems. Inor software failures. In most high availability
the event of a system failure,clustering ensures highconfigurations, nodes are isolated from each other so
availability to users. Access to mission critical data isa failure at one node does not affect the entire
not lost.system. In such a case, surviving nodes compensate
Redundant hardware components such as additionalfor the loss of the failed node through recovery
nodes,interconnects and disks allow the cluster toandthe system continues to provide data access to
provide high availability. Such redundant hardwareusers. This means data is consistentlyavailable, more
architectures avoid single points-of-failure and provideso than it would be with a single node upon node
exceptional fault resilience.failure. High availability also implies increased database
With Real Application Clusters,we de-couple theavailability.
Oracle Instance (the processes and memoryTransparency
structures running on a server to allow access to theThe concept of transparency is the functional
data) from the Oracle database(the physicalequivalent of single instance Oracle and shared
structures residing on storage which actually hold theconfigurations that use Real Application Clusters.
data,commonly known as data files).Applications that run on single instance Oracle
A clustered database is a single database that can beexecute with the same results using Real Application
accessed by multiple instances. Each instance runs onClusters. An Oracle database can be configured to
a separate server in the cluster. When additionalexecute in three different modes:
resources are required, additional nodes and instances• Single instance exclusive
can be easily added to the cluster with no downtime.• Shared with a single instance
Once the new instance is started, application using• Shared with two or more instances
services can immediately take advantages of it withBuffer Cache Management
no changes to the application or application server.Within a single instance, Oracle stores resources, such
Oracle Clusterwareas data block information, in a buffer cache that
Oracle provides Oracle Clusterware, a portableresides in memory. Storing this information locally
clusterware solution that is integrated and designedreduces the disk Input/Output (I/O) necessary for
specifically for Oracle Database. You no longer havedatabase operations. Since each node in Real
to purchase third party clusterware in order to haveApplication Clusters has its own memory that is not
a RAC database. Oracle Clusterware monitors andshared with other nodes, Real Application Clusters
manages Real Application Cluster databases. When amust coordinate the buffer caches of different
node in the cluster is started, all instances,listenersnodes while minimizing additional disk I/O that could
and services are automatically started. If an instancereduce performance. The Oracle Global Cache Service
fails the clusterware will automatically restart thetechnology maintains the high-performance features
instance so the service is often restored before theof Oracle while coordinating multiple buffer caches.
administer notices it was down.Fast Commits, Group Commits, and Deferred Writes
Hardware ArchitectureFast commits, group commits, and deferred writes
Oracle Real Application Clusters is shared everythingoperate on each instance in Oracle and work the
architecture. All servers in the cluster must share allsame whether in exclusive or shared mode. Oracle
storage used for a RAC database. The type of diskonly reads data blocks from disk if they are not
storage used can be network attachedalready in the buffer caches of one of the instances.
storage(NAS),storage area network(SAN), or SCSIBecause data block writes are deferred, they often
disk. Your storage choice is dictated by the servercontain modifications from multiple transactions.
hardware choice and what your hardware vendorOptimally, Oracle writes modified data blocks to disk
supports. Choosing a storage system that will provideonly when necessary:
scalable I/O for your application an I/O system that• When the blocks have not been used recently
will scale as additional servers are added to theand new data requires buffer cache space (in shared
cluster.or exclusive mode)
A Cluster requires an additional network to the Local• During checkpoints (in shared or exclusive mount
Area Network (LAN) that a database server ismode)
attached to for application connections. A cluster• When another instance needs the blocks (only in
require a second server is attached to for applicationshared mode)
connections. A cluster requires a second private• Cache copies of dirty blocks (changed blocks)
network commonly known as the interconnect.across the interconnect (write/write consistency).
Oracle recommends that you use 2 networkRow Locking and Multi version Read Consistency
interfaces for this network for high availabilityOracle’s row locking feature allows multiple
purposes. A network interface bonding external totransactions from separate nodes to lock and update
Oracle should be used to provide fail over and loaddifferent rows of the same data block. This is done
balancing. The interconnect is used by cluster forwithout any of the transactions waiting for the
inter-node messaging. The interconnect is also usedothers to commit. If a row has been modified but
by RAC to implement the cache fusion technology.not yet committed, then the original row values are
Virtual Internet Protocol Address (VIP)available to all instances for read access. This is called
Oracle RAC 10g requires a virtual IP address for eachmulti version read consistency.
server in the cluster. The virtual IP address is an