Databases Reference
In-Depth Information
Cache Fusion means that you do not have to work around the problems of pinging.
With Real Application Clusters, you will be able to see significant scalability improve‐
ments for most all applications, without any modifications. With that said, for OLTP
applications deployed to RAC (where there are frequent modifications to indexes within
a small set of leaf blocks), reverse key indexes might be used to distribute inserts across
leaf keys in the index and eliminate possible performance issues for this special situation
(see
Chapter 4
for an explanation of reverse key indexes).
Real Application Clusters also delivers all the availability advantages that were a part of
OPS. Because all the machines in a Real Application Cluster share the same disk, the
failure of a single machine does not mean that the database as a whole has failed. The
users connected to the failed machine have to be failed over to another machine in the
cluster, but the database server itself will continue to operate.
As of Oracle Database 10
g
, the model implemented with RAC was extended beyond
clusters to grid computing. Oracle now offers all the components you need to use to
implement clusters on several operating system platforms as part of the Oracle software
stack, including a volume manager and clusterware. In Oracle 10
g
Release 2, Oracle
made it possible to monitor the different nodes in a cluster and to issue advisories to
ensure better load balancing across the nodes.
Oracle Database 11
g
Release 2 added the concept of
server pools
, which provide a higher
level of organization within a RAC cluster. A server pool can have multiple nodes within
it, and you can designate an application to run on all instances within the pool or a single
instance. Instances can be added to a pool based on defined policies to handle shifting
workload requirements.
Exadata and OLTP
The Oracle Exadata Database Machine and the Exadata Storage Server software repre‐
sented a significant enhancement to the overall Oracle Database story, delivering per‐
formance that was up to an order of magnitude better than any other platform for the
Oracle Database. Exadata is discussed in more detail elsewhere in this topic.
In the first release, Exadata, the Exadata Storage Server software, was primarily used to
speed up data warehouse operations, since the most dramatic improvements in perfor‐
mance revolved around read operations for large numbers of rows, which are not that
common in OLTP workloads.
With the release of the Exadata X3 models and corresponding Exadata Storage Server
software in 2012, a new feature called Smart Flash Cache
writeback
was added. Write‐
back takes advantage of the speed of I/O to the Flash Cache in each Exadata Storage
Cell. With writeback enabled, write operations to disk are sent to the Flash Cache instead
of the much slower disk writes. This option eliminates the overhead of random disk
