Oracle X6-2-HA Deployment And User's Manual page 462

Consider the space requirements for your database and the policy for local disk
backups versus external backups. Typically, external backups have more space
available for the database than local backups.
• Container databases are created on Oracle ACFS.
• The log file size assumes four redo log groups for each instance with a log switch
every 15 minutes when the system is running at full capacity.
• Storage is shared between the servers on Oracle Database Appliance.
About Oracle Database Appliance Template Options
The database templates are configured specifically for the type of database workload
that you want to carry out on your databases on Oracle Database Appliance. Choose
the template that best matches the common workload your databases perform (OLTP,
DSS, In-Memory).
The database sizing tables provide template names and sizing based on the number
of CPUs and memory attributes for each type of database workload.
Identify the template type that is appropriate to your database workload and hardware:
• Use Oracle Database Appliance OLTP Database Templates if your database
workload is primarily online transaction processing (OLTP).
• Use Oracle Database Appliance DSS database templates if your database
workload is primarily decision support services (DSS) or data warehousing.
• Use Oracle Database Appliance In-Memory (IMDB) database templates if your
database workload can fit in memory, and can benefit from in-memory
performance capabilities.
• Use the platform-specific database templates if your database is on a specific
platform. For example, use Oracle Database Appliance X6-2-HA generic database
templates if your database is on Oracle Database Appliance X6-2-HA.
The templates provide general guidelines about the maximum number of databases
on a fully deployed Oracle Database Appliance. Base the memory and CPU sizing on
the number of instances per server, not the number of databases. Memory and CPUs
are shared by all databases and the number of databases impacts performance. For
example, when you have one database, the database receives full I/O throughput. If
you have 9 databases, then each database has only 1/9th of the available I/O
throughput for all disks.
When you have multiple databases, they share and compete over available CPU
resources. When choosing a template, do not have more CPU_COUNTs distributed
than available threads in the system. Oracle recommends partitioning, where every
database has exclusive CPUs. If you choose to exceed the recommended maximum
number of databases on Oracle Database Appliance (over-provision), be aware of
your I/O results and performance degradation. With the over-provisioned approach,
the sum of the CPU_COUNT across all instances can exceed the number of CPUs.
Better resource utilization is provided with over-provisioning; however, it is possible for
contention to occur when multiple databases are heavily loaded at the same time.
Over-provisioning is best used for systems running non-critical applications. Avoid
using over-provisioning for databases with high I/O requirements or those with high
transaction rates. If you choose to over-provision, then do not exceed twice the
number of total CPUs. Twice the allocation of total CPUs is based on hyper-threading
of 2 CPU threads per core.
About Database Templates for Oracle Database Appliance
Appendix E
E-2