Capacity Planning and Architecture - Expert Oracle RAC Performance Diagnostics and Tuning - page 36

As listed in the preceding, the LIOT is set to 200 seconds, and SIOT is set to 27 seconds. LIOT is used to determine
the disk write latency, and SIOT is used by CSS during cluster reconfiguration. SIOT is similar to the misscount value
used by the NHB; however, it is computed based on the reboot time (default reboot time is set to 3 seconds). The
disktimeout , misscount , and reboottime can be determined using the following crsctl command:
[root@ssky1l4p1 orarootagent_root]# crsctl get css disktimeout
CRS-4678: Successful get disktimeout 200 for Cluster Synchronization Services.
[root@ssky1l4p1 orarootagent_root]# crsctl get css reboottime
CRS-4678: Successful get reboottime 3 for Cluster Synchronization Services.
[root@ssky1l4p1 orarootagent_root]# crsctl get css misscount
CRS-4678: Successful get misscount 30 for Cluster Synchronization Services.
Based on the default values listed in the preceding, the SIOT is 27 seconds ( misscount less reboottime ).
Local Heartbeat (LHB)
LHB is an internal heartbeat mechanism where the message is sent to the cssdmonitor and the cssdagent to keep
them informed about the health of the CSS. LHB notifications also happen every second and use and share the same
thread with the NHB and DHB.
System Change Number (SCN)
The SCN is required in a single instance configuration to serialize activities such as a block changes, redo entries, and
replay of redo logs during a recovery operation. It has a more robust role in a RAC environment.
In a RAC configuration, more than one instance can make updates to the data blocks. These data blocks are
transferred via the cluster interconnect between the instances. To track these successive generations of data blocks
across instances, Oracle assigns (uses), to each data block that is generated, a unique logical timestamp or SCN. The
SCN is used by Oracle to order the data block change events within each instance and across all instances.
In a RAC environment, separate SCNs are generated by each instance. However, in an effort to keep the
transactions in a serial order, these instances have to resynchronize their SCNs to the highest SCN known in the cluster.
The method used by Oracle to synchronize its SCN to the highest SCN in the cluster is called broadcast on
commit . Under this method, SCNs are propagated to other instances when a data is committed on an instance,
meaning Oracle does not wait to piggyback the SCN change on to another message. Broadcast on commit is
implemented by reducing the default value defined by the parameter MAX_COMMIT_PROPAGATION_DELAY . Reducing the
value to less than 100 hundredths of a second increases the SCN propagation between instances.
Mastering of Resources
Based on the demand for resources on a specific file, the resource is maintained on the instance whose use of it is
highest: for example, if instance SSKY1 was accessing an object A1 and data from that object was being processed
for about 1,500 user requests all connected to instance SSKY1 ; and say instance SSKY2 also required access to the
object A1 for 100 users. It's obvious that SSKY1 has more users accessing this object A1. Hence, instance SSKY1 would
be allocated as the resource master for this file, and the global resource directory (GRD) for this object would be
maintained on instance SSKY1 . When instance SSKY2 requires information from this object, it must coordinate with
the GCS and the GRD on instance SSKY1 to retrieve/transfer data across the cluster interconnect.