Monitoring the Hosting Platform - Expert Consolidation in Oracle Database 12c - page 277

• Medium configuration : This is a bit more realistic featuring 2 Management Hosts, between
1000 and 10,000 targets monitored by a 100 to a maximum of 1000 agents. There are an
expected 10-25 concurrent user sessions.

• Large configuration : The term “large configuration” is used for anything more than the
medium configuration, but does not exceed 50 concurrent users.

Depending on the configuration type you need different hardware. Since the OMS is written mainly in Java
and executed in a WebLogic environment, it makes sense to be generous with RAM. The recommendations made
by Oracle in the installation guide are probably too conservative. The suggestions are to use 8GB, 12GB, and 20GB
RAM for small, medium, and large configurations respectively. The rule of thumb here is that RAM can only be
replaced with more RAM! If you can spare the expense to upgrade the host to more than that by all means do it.
Remember from the hardware chapter that it is easily possible to add more than half a TB of RAM into a dual-socket
Xeon E5 system.
Related to the overall amount of memory available to the system you should consider increasing the amount of
memory available to the WebLogic process. Medium-sized deployments should set the heap size of the Java processes
on the OMS to 4 GB. Oracle recommends up to 8 GB heap size for large deployments.
Hard disk space is another consideration to be made. If you are planning to patch databases or otherwise
download and stage software via the Self-Update functionality, you need to reserve space for the software library. In
environments where you are using more than one OMS, the software library has to be in a sharable location. For UNIX
that most often means NFS.

Considerations for the Repository Database
The management repository is the other main part of the infrastructure. Depending on the level of availability
expected from the Enterprise Manager deployment, the repository database needs to be protected from instance
failures. Please refer back to Chapter 4 for more information about possible ways to protect the database. The
Maximum-Availability-Architecture (MAA) documents suggest the use of the Real Applications Cluster option for the
management repository.
The suggestions for the repository database servers are again conservative by today's industry standards. Using
the same classes of configuration as described in the previous section, the recommendation for CPU cores is to use
2/4/8 for small/medium/large configurations and 6/8/16 GB RAM. These can only be the bare minimum; smooth
operations of a sizable deployment realistically require more capacity. Luckily even a two-socket x86-64 can support
up to 16 cores with Intel or 32 modules with the current AMD processors.
The repository database stores the information sent by the management agents in three tablespaces created
during the installation:

• MGMT_ECM_DEPOT_TS

• MGMT_TABLESPACE

• MGMT_AD4J_TS

Depending on how aggressively you decide to purge history, you need to take a considerable amount of space
into account for these tablespaces. The minimum recommended space for a large configuration is approximately
400 GB.
Part of the planning for the repository database should be a check of the certification matrix. Not every database
release is certified for use as a repository database! Enterprise Manager makes some assumptions about the
database initialization parameters in addition to the above-mentioned tablespaces. Recommended initialization
parameters and other important information with regards to the repository database are listed in Chapter 11 “Sizing
Your Enterprise Manager Deployment” of the “Cloud Control Advanced Installation and Configuration Guide” for
Enterprise Manager 12c Release 2.