Many factors can affect the resources used by Oracle Enterprise Manager Ops Center, and ultimately its ability to scale to manage large deployments.
The following topics are covered:
You can scale Oracle Enterprise Manager Ops Center to accommodate a variety of environments. However, the size and layout of your environment can effect the number of Enterprise Controllers and Proxy Controllers that you require. The scaling guidelines provided here are based on the use of an external repository and are minimum required parameters.
For sizing purposes, the following are defined as assets:
Servers, including the service processor and operating system
Switches
Storage devices
Non-global zones
Logical domains
Chassis
Oracle SPARC Enterprise M-Series chassis
Oracle SPARC Enterprise M-Series domains
Oracle Exadata Storage Server X2-2 cells
See Determine the Number of Asset Access Points for how to run a script on the Enterprise Controller to determine the number and type of assets on each Proxy Controller and the total number of assets on the Enterprise Controller.
Consider the following factors when selecting a platform for the Enterprise Controller:
System: The Enterprise Controller has many parallel executing processes and a multi-core system will significantly improve performance. There are some processes that are by their nature single threaded and benefit from higher clock speed. It is recommended to choose a process that is both multi -core and has a high single thread CPU speed.
Note:
Do not install the Enterprise Controller on a hardened system. The installation will complete successfully, but you will encounter several issues when you use the software.Memory: The core JVM machine consumes memory and its resource utilization will increase with the number of assets being managed. If the Java virtual machine is forced into virtual memory, then Enterprise Manager Ops Center performance will reduce significantly.
Database: A database is present in Ops Center and a high performing disk I/O is important in the implementation. Generally, disk striping or the use of a well configured high performing file system with such features built in, such as ZFS, is highly recommended. See Management Repository Considerations for information on database sizing.
Operating system: While not a scaling issue, the system must be running on a version of Oracle Solaris 11 if you plan on using Oracle Enterprise Manager Ops Center to manage Oracle Solaris 11 Operating Systems or virtualized environments, or if you plan on discovering and managing Oracle SuperCluster and other Oracle engineered systems. Because Oracle SuperCluster uses Oracle Solaris 11, you must install your Enterprise Controller on Oracle Solaris 11 if you plan to manage Oracle SuperCluster.
Note:
It is recommended that you use Oracle Solaris 11.2.Network: Start-up times of the Enterprise Controller can be minimized by ensuring that the network link to the repository is as low latency as possible. The start-up process involves a large number of queries as the state machine is being built. It is recommended that a 10 GigE connection be available between the Enterprise Controller and Management Repository.
For a remote customer-managed database, set up the network according to the Oracle database networking setup defined in the Oracle Server Installation Guide. See the Oracle Database Enterprise Edition documentation at http://www.oracle.com/technetwork/database/enterprise-edition/documentation/index.html for detailed database requirements. See Management Repository Considerations for more information on database considerations.
An Enterprise Controller can manage up to 3,000 mixed assets including hardware, operating systems, virtualization technologies, and other assets. Deployments above 1,000 assets might require you to tune your Java Virtual Machine (JVM) heap size.
An Enterprise Controller system must meet the following minimum requirements:
Oracle SPARC or x86 platform that is running a supported Oracle Solaris or Linux OS, such as Oracle Solaris 11
87 GB of disk space
Swap space
Embedded (Ops Center local) database: determined by the amount of RAM (maximum 16 GB of swap space is required)
If RAM is between 8 GB and 16 GB, then match RAM
16 GB for RAM greater than 16 GB, then use .75 times RAM
Customer-managed (remote) database: 6 GB Swap Space
8 GB of Java Heap Size, as indicated by the java flag -Xmx
Consider the following factors when selecting a platform and environment for the Proxy Controller:
Operating system: To provision and update Oracle Solaris 11 operating systems, the Enterprise Controller and Proxy Controller must be installed on at least Oracle Solaris 11.1 operating system.
Global or non-global zone: If the assets being managed are Oracle Solaris 11, you can configure the Proxy Controllers in a global or non-global zone. Otherwise, install the Proxy Controllers in global zones. A Proxy Controller includes the capability to perform OS provisioning, which is not supported by Oracle Solaris 10 non-global zones.
Network I/O: If the Proxy Controller is supporting Operating System provisioning then it is recommended that the Proxy Controller should be utilizing GigE network cards or higher.
Consider installing additional Proxy Controllers, beyond the minimum requirements, in environments with the following characteristics:
Remote datacenters: If some of your assets are in a remote location, installing a Proxy Controller in that location can reduce network latency and cache Oracle Enterprise Manager Ops Center content closer to the assets.
Deployment through firewalls: If you have firewalls between your Enterprise Controller and some managed assets, having a Proxy Controller near the assets can simplify deployment and management.
High Availability: Installing additional Proxy Controllers gives you the ability to migrate assets between Proxy Controllers for maintenance or if a Proxy Controller fails.
Virtualization: If you intend to manage virtualization technologies like Oracle VM Servers for SPARC, additional Proxy Controllers help handle the added asset and job load.
The software is supported on a wide variety of hardware, operating systems, and virtualization technologies. See the Oracle Enterprise Manager Ops Center Certified Systems Matrix for a list of certified systems.
Each Proxy Controller can manage up to 450 assets for a basic environment and up to 350 assets for a more complex environment. Installing additional Proxy Controllers beyond the requirements is beneficial in some environments.
A Proxy Controller system must meet the following minimum requirements:
Oracle SPARC or x86 platform that is running a supported Oracle Solaris or Linux OS, such as Oracle Solaris 11
1.5 GB of disk space
6 GB Swap space
2 GB Java Heap size, defined by -Xmx in the java-flags
It is generally recommended that you use proxy fan out as it makes sense in the network to be deployed. For example, OS provisioning capabilities are often restricted to a network subnet.
Changing the resource configuration of the system will affect the numbers, as will more complex mixing of the services across the deployment.
An agent on an Oracle Solaris Zone Virtualization Controller installed on a system with one (1) processor and 128 GB of memory can manage with a good level of performance:
50 non-global zones with one (1) filesystem configured on each non-global zone.
15 non-global zones with 20 filesystems configured on each non-global zone.
An agent on an Oracle VM Server for SPARC Virtualization Controller installed on a system with one (1) processor and 128 GB of memory can manage, with a good level of performance, 20 logical domains with 25 disks (LUNs) configured on each domain.
For both configurations described above, the agent deployment time will take less than one hour and the restart time will be about 15 minutes. The CPU consumption of the agent should not exceed 5 percent.
Oracle Enterprise Manager Ops Center stores product data in an Oracle Database Enterprise Edition database. This database can be an embedded database installed on the same system as the Enterprise Controller, or a customer-managed database that is accessible to the Enterprise Controller system. Run the OCDoctor to check for basic database requirements.
A backup feature enables you to back up your environment. The backup feature supports an embedded or customer-managed database. If you are using a customer-managed database, you can also back up the database schema using the existing backup and recovery processes implemented by your database administrator.
Note:
If you plan to implement high availability for the Enterprise Controller, you must use a customer-managed database.Use the following sizing guidelines to estimate the maximum size of the database based upon the number of assets managed:
Operational Data: 74 megabytes (MB) per asset.
Reporting Data: 166 MB per asset. Steady state is reached at 180 days when using the default rollup schedule. USERS tablespace (Reporting) is required when report attributes are enabled. Use one of the following methods to change the size of the reporting data:
Customize the amount of attributes shown in the Enterprise Controller.
Stop the report service, which will stop generating report data in the database.
See Appendix A for more information about the report service.
UCE Data: 4 gigabytes (GB) needed for 10 channels. 12 GB additional for disconnected mode. You will generally need 400 megabits (Mb) for each additional channel. In some cases, you might need up to 600 Mb for a channel.
Oracle Solaris and Linux operating system updates and provisioning use UCE data. A channel, or UCE channel, is a supported Oracle Solaris or Linux operating system distribution, such as an Oracle Solaris 10 on a SPARC platform.
Schema Backups: Each backup dump requires 40% of Enterprise Manager Ops Center Schema Size on the database server in the OC_DUMP_DIR directory location.
Embedded, or Ops Center Local, database: A maximum of seven (7) copies of the backup are kept online.
Customer managed database: The database administrator must manage the Oracle Enterprise Manager Ops Center dump directory backup content.
Archival: An embedded database does not run in archive mode. For a remote database running in archive mode, the size of the archive is set according to the redo logs size.
Table 3-1 shows the estimated database size for when 1000 assets and 100 assets are managed and 10 UCE channels are configured in connected mode.
Table 3-1 Estimated Database Size
| Use Pattern | Size for 1000 Assets (MB) | Size for 100 Assets (MB) |
|---|---|---|
|
SYSTEM tablespace |
1000 |
1000 |
|
UNDO tablespace |
2750 |
275 |
|
SYSAUX |
1200 |
1200 |
|
USERS tablespace (Operational) |
100000 (100 GB) |
10000 (10 GB) |
|
USERS tablespace (Reporting) |
230000 (230 GB) |
23000 (23 GB) |
|
USERS tablespace (UCE) |
4000 |
4000 |
|
TEMP |
2000 |
200 |
|
TOTAL DB Size |
340950 |
39675 |
|
Spaced Reserved for Backups |
954660 |
103600 |
The database size also depends on your retention policy of gathered data. The default retention is 180 days for a 24 hour sample.
When using an embedded (Ops Center local) database, the Enterprise Controller system must meet the requirements for an Oracle Database Enterprise Edition installation. You can improve the embedded database performance by using the following configuration:
Use the ZFS filesystem as your underlying database
Set the redo log files to a block size of 8K
Set the logbias=throughput
Separating the redo logs onto a separate zpool (a different underlying disk) will further improve the embedded database performance.
Dynamic Intimate Shared Memory (DISM) allows a database to dynamically extend or reduce the size of the shared data segment. The embedded database is configured to use a maximum of 8 GB RAM using DISM (TOTALMEMORY in the response file.)
See the Oracle Enterprise Manager Ops Center Installation Guide for Oracle Solaris Operating System Guide or the Oracle Enterprise Manager Ops Center Installation Guide for Linux Operating Systems Guide for information about local database requirements.
A customer-managed (remote) database system must use a supported Oracle Database version. See the Oracle Enterprise Manager Ops Center Certified Systems Matrix for a list of supported database versions.For a reference supported DB versions are documented in Certified Systems.
The network connection to the remote database should have a low latency.
Note:
If you are implementing high availability, the remote database system must be accessible for all of the Enterprise Controller nodes.The recommended database memory size is 8 GB.
The following are the minimum initialization parameters to use when you create the database:
open_cursors=1000
pga_aggregate_target=0
processes=500
recyclebin=OFF
sga_target=0
See the Oracle Database Enterprise Edition documentation at http://www.oracle.com/technetwork/database/enterprise-edition/documentation/index.html for detailed database requirements.
When using a remote database, the network connection to the remote database should have a low latency.
For Ethernet connectivity, it is recommended that the network you are planning to use to conduct OS provisioning across be 1/10GB capable.
If your site uses VLAN, a best practice is to create a separate VLAN for management networks.
In a mirror setup, you need two (2) disks with following minimum disk space:
In the active and running boot environment:
/var: 85GB for the Enterprise Controller and Proxy Controller run time
/: 30GB for the Enterprise Controller and Proxy Controller install and data
Enterprise Controller Libraries
15 GB for the EC library
45 GB for the Oracle Solaris 11 Software Update Library
5 GB per channel for the Linux, Solaris 8-10 Software Update Library
Database
See Table 3-1 for database sizing. The following is an example estimated database size for 400 assets: 160 GB for the database plus 400 GB reserved for backups
Sufficient storage disk space is critical in optimizing performance.
The following configurations are applicable for a SPARC platform:
The OS disk configuration is on a Fibre Channel hardware RAID.
The Oracle Enterprise Manager Ops Center data is on a Fibre Channel hardware RAID.
The following are recommended disk configurations:
Place the /var/opt/sun/xvm directory on a different set of disk spindles from the operating system. Do not co-mingle with the OS.
Use at least 4 spindles in the hardware RAID cache configuration.
Place the OS on multiple spindles with hardware RAID cache. If the hardware does not allow this with internal drives, then boot from SAN
The Proxy Controller software is considered relatively lightweight and, in all but the extreme scale cases, does not require a dedicated system.
The following configurations are applicable for SPARC platforms:
The OS disk configuration is on a Fibre Channel hardware RAID.
The Oracle Enterprise Manager Ops Center data is on a Fibre Channel hardware RAID.
An access point is a connection between a managed asset and the Enterprise Controller through a Proxy Controller. The number of access points for the Enterprise Controller and for each Proxy Controller provides more accurate scaling information than the number of assets.
A single managed asset can have one or more access points, depending on how many Proxy Controllers are managing it. For example, if an Oracle VM Server for SPARC control domain's operating system is managed by one Proxy Controller while a hosted logical domain's operating system is managed by another Proxy Controller, the Enterprise Controller has two access points for that asset.
The OCDoctor utility includes a script, AssetCount.sh, that you can run on the Enterprise Controller to get the current access point count for the Enterprise Controller and each Proxy Controller, including a breakdown of the type of assets on each Proxy Controller.
When the OCDoctor utility is run with the --troubleshoot option, it performs an access point count. If the results of this count indicate that the Enterprise Controller or Proxy Controllers are approaching or beyond the maximum recommended number of assets, the output will include a warning.
When you run the ./OCDoctor.sh --collectlogs command on the Enterprise Controller, the current access point count is included in the output. The current access point count can be useful when troubleshooting an issue.
The following methods can be used to view the access point count:
You can view the total number of access points managed by the Enterprise Controller, as well as a categorized count of the access points managed by each Proxy Controller. Run the following command on the Enterprise Controller:
# /var/opt/sun/xvm/OCDoctor/toolbox/AssetCount.sh standard
For example:
# /var/opt/sun/xvm/OCDoctor/toolbox/AssetCount.sh standard EC 72 Proxy Assets Zones Ldoms OVMGuests Servers Storages Switches ExadataCells MSeriesChassis MSeriesD ------------------------------------------------------------------------------------------------- pc4 32 5 25 0 2 0 0 0 0 0 pc1 28 0 26 0 2 0 0 0 0 0 pc0 12 2 4 0 6 0 0 0 0 0 Use option '-v' to see additional notes on the counting algorithm.
You can view a list of the access points of each type that are managed by each Proxy Controller. The output first provides the total number of access points on the Enterprise Controller, then for each Proxy Controller provides the total number of access points, the number of access points of each type, and the user-friendly name for each asset.
The output for this command can be very long in a large environment.
Run the following command on the Enterprise Controller:
# /var/opt/sun/xvm/OCDoctor/toolbox/AssetCount.sh machine
For example:
# /var/opt/sun/xvm/OCDoctor/toolbox/AssetCount.sh machine EC 72 Proxy 32 pc4 Zones 5 S11zone101 S11zone102 S11zone100 S11zone103 S11zone104 Ldoms 25 stdldom21 stdldom34 stdldom36 stdldom22 stdldom45 stdldom47 ... OVMGuests 0 Servers 2 10.187.70.169 pc4 ... Proxy 28 pc1 Zones 0 Ldoms 26 stdldom21 stdldom34 stdldom36 stdldom22 stdldom45 stdldom47 ... OVMGuests 0 Servers 2 10.187.70.171 pc1
You can view a count of the access points for each Proxy Controller in three categories: Agent Controllers, Agentless assets, and service processors.
Run the following command on the Enterprise Controller:
# /var/opt/sun/xvm/OCDoctor/toolbox/AssetCount.sh agent
For example:
# /var/opt/sun/xvm/OCDoctor/toolbox/AssetCount.sh agent EC 72 Proxy Agents Agentless SPs -------------------------- pc4 25 2 0 pc1 1 1 0 pc0 5 5 5 Use option '-v' to see additional notes on the counting algorithm.
You can view all access point data, including the results of the standard, machine, and agent options.
Run the following command on the Enterprise Controller:
# /var/opt/sun/xvm/OCDoctor/toolbox/AssetCount.sh all
For example:
# /var/opt/sun/xvm/OCDoctor/toolbox/AssetCount.sh all EC 72 Proxy Assets Zones Ldoms OVMGuests Servers Storages Switches ExadataCells MSeriesChassis MSeriesD ------------------------------------------------------------------------------------------------- pc4 32 5 25 0 2 0 0 0 0 0 pc1 28 0 26 0 2 0 0 0 0 0 pc0 12 2 4 0 6 0 0 0 0 0 Proxy Agents Agentless SPs -------------------------- pc4 25 2 0 pc1 1 1 0 pc0 5 5 5 Proxy 32 pc4 Zones 5 S11zone101 S11zone102 S11zone100 S11zone103 S11zone104 Ldoms 25 stdldom21 stdldom34 stdldom36 stdldom22 stdldom45 stdldom47 ... OVMGuests 0 Servers 2 10.187.70.169 pc4 ... Proxy 28 pc1 Zones 0 Ldoms 26 stdldom21 stdldom34 stdldom36 stdldom22 stdldom45 stdldom47 ... OVMGuests 0 Servers 2 10.187.70.171 pc1 Use option '-v' to see additional notes on the counting algorithm.
See the Oracle Enterprise Manager Ops Center Installation Guides in the documentation library, which is located at http://docs.oracle.com/cd/E40871_01/index.htm.
See the Oracle Enterprise Manager Ops Center Administration Guide for more information about the OCDoctor, including tuning and troubleshooting options.OCDoctor will print a WARNING if the Asset Count of an EC or Proxy is reaching the maximum recommended value, or is even beyond it
See the Oracle Enterprise Manager Ops Center Certified Systems Matrix for a list of supported Oracle Database versions.
See the Oracle Database Enterprise Edition documentation at http://www.oracle.com/technetwork/database/enterprise-edition/documentation/index.html for detailed database requirements.