Capacity Planning and Architecture - Expert Oracle RAC Performance Diagnostics and Tuning - page 33

NM alternates its actions between the heartbeat check using the interconnect called Network Heartbeat (NHB)
and the heartbeat to the voting disk called Disk Heartbeat (DHB) to determine the availability of other nodes in the
cluster. (See the “Heartbeats” section later in this chapter.)
The GM provides group membership services. All clients that perform I/O operations register with the GM; for
example, the LMON , DBWR , and so forth. Reconfiguration of instances (when an instance joins or leaves the cluster)
happens through the GM. When a node fails, the GM sends out messages to other instances regarding the status.
Event manager daemon or EVMD is an event forwarding daemon process that propagates events through
the Oracle Notification Service (ONS). It also scans the node callout directory and invokes callouts in reactions to
detected events, for example, node up and node down events. Although this daemon is started subsequent to the
OCSSD, EVMD is the communication bridge between the Cluster ready service daemon (CRSD) and OCSSD. All
communications between the CRS and CSS happen via the EVMD.
CRSD or Oracle Clusterware daemon function is to define and manage the resources. A resource is a named
entity whose availability is managed by the Clusterware. Resources have profiles that define metadata about
them. This metadata is stored in the Oracle Cluster Registry (OCR). CRS reads the OCR. The daemon manages the
application resources: starts, stops, and manages failover of application resources; generates events during cluster
state changes; and maintains configuration profiles in the OCR. If the daemon fails, it automatically restarts. The OCR
information is cached inside the CRS. Beyond performing all the functions, CRS also starts and communicates with
the RACGIMON daemon process.
Resources that are managed by the CRS include global service daemon (GSD), ONS daemon, virtual Internet
Protocol (VIP), listeners, databases, instances, and services. Resources are grouped based on the level at which they
apply to the environment. For example, some of these resources are referred to as node applications ( nodeapps ),
and they pertain to individual nodes in the cluster. These resources are needed on a per node basis independent of
the number of databases on the node. GSD, ONS, VIPs, and listeners are the list of nodeapps . They are created and
registered with the OCR during installation of the Oracle Clusterware. Listener, database, and service resources are
created during the database creation process.
Cache Fusion
Cache fusion is a technology that uses a high-speed IPC to provide cache-to-cache transfer of data blocks between
instances in a cluster. This technology for transferring data across nodes through the interconnect became a
viable option as the bandwidth for interconnects increased and the transport mechanism improved. Cache fusion
architecture is revolutionary in an industry sense because it treats the entire physical distinct RAM for each cluster
node logically as one large database system global area (SGA), with the interconnect providing the physical transport
among them.
The global cache service (GCS) and global enqueue service (GES) processes on each node manages the cache
synchronization by using the cluster interconnect. Cache fusion addresses transaction concurrency between
instances. The different scenarios of block sharing are broadly stated as the following:
• Concurrent reads on multiple nodes : This occurs when two or more instances participating
in the clustered configuration are required to read the same data block. The block is shared
between instances via the cluster interconnect. The first instance that reads the block would
be the owning instance, and the subsequent instances that require access to the same block
will request it via the cluster interconnect.
• Concurrent reads and writes on different nodes : This is a mixture of read/write operations
against a single data block. A block available on any of the participating instances could be
modified by a different instance while maintaining a copy/image that is different from the
database. Such transactions use the interconnect. A block can be read as is, that is, in a current
version, or a read consistent version could be built by applying required undo.
• Concurrent writes on different nodes : This is a situation in which multiple instances request
modification of the same data block frequently.