Professional Documents
Culture Documents
Users
Database
How RAC clustering is done
One set of data
All nodes in the cluster see the same set of data
All nodes have access
Any node can update the data
Increased Manageability
One virtual system to configure and manage
Single Oracle Database
Single management console
Single system image for the database integrated with the
cluster
Cluster-wide monitoring and diagnostics
Oracle Enterprise Manager Integration (9i)
Oracle DBConsole and Grid Control Integration
(10g)
What’s shared; What’s not
Shared
Disk access
Resources that manage data
All instances have common data & controls files
Not Shared
Each node has its own dedicated:
System memory
Operating system
Database instance
Application software
Each instance has individual
Log files and
Rollback segments
RAC can perform
Load-balancing
Failover
Load-Balancing through the Listener
Client
r Lis
ne te
r
e ne
ne
Lis
Lis r
te
t
s
en
Li
er
Node 1 Node 2 Node 3 Node 4
Database
How workload is balanced
Nodes report CPU
usage to listeners
Node 1
Database
Client
Node 2
How workload is balanced
Listeners choose least
busy node when
request comes in from
client Node 1
Client Database
Node 2
Load-Balancing
Users
Database
Failover
Users
X
Server Server Server Server
Database
Transparent Application Failover
Masks failures to end users; they don’t need to log
back into the system
Applications and users are transparently reconnected
to another node
Applications and queries continue uninterrupted
Transactions can failover and replay
Login context maintained
DML transactions are rolled back
RAC Improvements for Oracle 9i
Full Cache Fusion
Enhanced coordination of cache management and
distributed lock manager (DLM)
Lock simplification and automation
Global Cache Service coordinates local buffer cache
and remote block transfers
Enhanced IPC (InterProcess Communication)
Resource configuration simplification and automation
Oracle 10g RAC New Features
Integrated Clusterware Management
No third-party clusterware software required
Automatic Workload Management
Application workloads can be managed through named services
Single System Image Management
Enterprise Manager manages RAC instances as a single image
Fast Connection Failover
Fast recovery between the database and mid-tier applications
Performance Improvements
Reduced message traffic, memory usage, and other resources
Zero Downtime Patching
Patches may be applied one node at a time without downtime
Cluster Verification and Improved Diagnostic Tools
New cluster diagnostic tool and improved diagnostic tools
Full Cache Fusion
Is a major feature of RAC starting with 9i
The underlying technology that enables RAC
Protocol that allows instances to combine their data
caches into a shared global cache
Allows any node to get the most up-to-date data
information from the cache of any other node in the
cluster without having to access the disk drives
again.
Improved performance with 10g
What is Cache Fusion? When do I
care about it?
“Dirty” block of data is created
Data from disk is read into memory on a node
Data is updated on that node
Data hasn’t been written to disk yet.
Another node requests the data
“ABC” block of data written to the
disk drives in the database
Node A Node B
ABC Data
“ABC” block of data read into
memory on Node A
Node A Node B
ABC Data
ABC Data
“ABC” updated to “XYZ” in cache
Node A Node B
ABC Data
XYZ Data
ABC Data
Node B requests data block
I want
data!
Gimme!
Node A Node B Gimme!
ABC Data
XYZ Data
ABC Data
Node A must write data block to
disk drive
I want
data!
Gimme!
Node A Node B Gimme!
ABC Data
XYZ Data
A C
9 iR
t o
s
wr
o u
ite
v i
r e
ABC Data
XYZ Data P
Node B must read data block from
disk drive
Node A Node B
ABC Data
XYZ Data XYZ Data
A C
9 iR
t o
us
ad
io
re
e v
ABC Data
XYZ Data Pr
Now with RAC Cache Fusion
I want
data!
Gimme!
Node A Node B Gimme!
ABC Data
XYZ Data XYZ Data
ABC Data
Shared Cache Across Nodes
Users
Database
Resource Simplification and
Automation
No init.ora parameters required
Resource affinity
to move the location of the resource masters for a
database file to the instance where block operations are
most frequently occurring. This optimizes
Dynamic resource remastering
Ability to move the ownership of a resource between
instances of Real Application Clusters.
Dynamic resource remastering is used to implement
resource affinity for increased performance.
Review
What does cache fusion avoid that was mandatory
in previous versions of Oracle Parallel Server?
Which Oracle process is most important in
managing user session failover?
If the purpose of the interconnect is NOT to serve
as a “heartbeat”, where is the heartbeat?
Summary
New Features
Shared Everything Clustering
Cache Fusion
RAC Clustering failover & load-balancing