BMC Compact RDS performance benchmarks

This topic describes the tests that BMC conducted with the BMC Compact RDS 4.0 deployment to establish the performance benchmarks. The following information is included:

• • Performance testing objectives
  • Test environment
  • Test cases
    • Test case 1 – Determine end-to-end time to provision VMs
    • Test case 2 – Determine UI response time for 10 and 25 concurrent users
  • Test results
    • VM provisioning
    • UI response time
    • Longevity
  • Performance tuning parameters
    • VC BSA Agent
    • Oracle database tuning
    • Less-than-optimal VM cloning performance at ESX server or VC server requires additional tuning

Performance testing objectives

The key objectives of this performance testing exercise were:

• Determining end-end time for provisioning virtual machines (VMs)
• Determining UI response time for key cloud administrator and technical end user transactions
• Determining end-to-end time for VM provisioning with concurrent user UI transactions

Test environment

The test environment consisted of the following components:

• ESX server used for hosting BMC Compact RDS

  Model	CPU	Memory	Datastore	Version
  Dell R900	16@2.40 GHz	128 GB	Local 1.09 TB	5.0.0

• Target ESX server used for provisioning test VMs

  Model	CPU	Memory	Datastore	Version
  Dell R710	8@2.40 GHz	64 GB	Local 1.91 TB	4.1.0

• VMware Virtual Center (VC) server details (target infrastructure):
• vSphere version: 4.1.0
• VM template specification:
  • OS: Windows 2003 EE
  • Disk size: 5 GB
  • Disk provisioning: Thin
  • Disk location: Stored on local datastore of each ESX server

Test cases

The following primary test cases were considered for benchmarking:

Test case 1 – Determine end-to-end time to provision VMs

1. 
   
   • Single user—Single Tier: A user provisioning a single VM having only an OS installed on it
   • Single User—Multiple VM instances (bulk provisioning): A user submits a single request to provision multiple VMs, which can be done by specifying the number of VMs to be provisioned in the Quantity field on the Submit Request page.

Test case 2 – Determine UI response time for 10 and 25 concurrent users

Test case 2 determined the UI response time for 10 and 25 concurrent users, by performing the following key cloud user transactions. This test used the HTTP Watch tool on the Mozilla 23 browser.

1. Launch
2. Log on
3. Click CLM Admin Console
4. Search Location
5. Click Pod
6. Search Pod
7. Click Network Containers
8. Search Network Container
9. Click Compute Pool
10. Click Compute Pool List View
11. Search Compute pool
12. Click Resources
13. Refresh Resources
14. Search Virtual Cluster
15. Change Resource Type to Virtual Resources pool
16. Search Virtual Resource pool
17. Click Providers workspace
18. Click Network Quick Link
19. Click Storage Quick Link
20. Click Configuration Workspace
21. Click Service Governor Workspace
22. Click Compute Pools Quick Link
23. Click Virtual Disk Repository Pools Quick Link
24. Click Network Quick Link
25. Click Tenants Workspace
26. Click Service Blueprint Workspace
27. Refresh Service Blueprint Workspace
28. Click Service Catalog Workspace
29. Refresh Service Catalog Workspace
30. Click Service Instance Workspace
31. Refresh Service Instance Workspace
32. Search Service Instance
33. Click Dashboard Workspace
34. Log out

Test results

This section provides the results from the performance tests and describes how the results were measured.

VM provisioning

BMC carried out provisioning tests for single VM and multiple VMs:

• The single VM provisioning test was carried out for the Windows 2003 template of size 5 GB with the Thin Provisioning option.
• The bulk provisioning test captured the time for bulk quantities of 10, 25, and 50 VMs for a single request.

The following figure shows the time to provision 1, 10, 25, and 50 VMs at the empty DB load:

 
The following figure shows the time to provision 1, 10, 25, and 50 VMs at the 5K DB load.

UI response time

UI response time for various user activities for Cloud users was captured by using a load testing tool, Silk Performer 2010 R2. The following figure shows the response time for UI actions carried out for 10 and 25 concurrent Cloud users over the LAN (approximately 0 ms network latency).

Longevity

No resource utilization or functional issues were observed during the longevity test, which was run to evaluate the stability of the solution under load over a longer period. In this test, approximately 800 VMs were provisioned and decommissioned every day and the following operations were performed on provisioned VMs:

• Extend the date of a VM
• Stop VMs
• Start VMs
• Decommission VMs
• Provision VMs
• Resource Utilization Graphs

The test was executed and monitored for one week (total number of VMs provisioned and decommissioned was approximately 5,500).

The following figures show the resource utilization of the BMC Compact RDS server during the longevity test.

System CPU utilization
 

System memory utilization
 

Performance tuning parameters

BMC Compact RDS is tuned for optimal performance. All the required tuning for each product is already configured in BMC Compact RDS. The VMware Virtual Center (VC) server for managing target infrastructure is an external component for BMC Compact RDS. This section describes the recommended parameter changes on the VC server for optimal performance.

VC BSA Agent

BMC Compact RDS solution post-installation configurations require the VC to configure. Perform the following configurations at the VC level.

1. Change the following parameters listed in the AssetImplConfig.xml file located at:
   AgentHomeDirectory\8.1\RSCD\daal\Implementation\BMC_VMware_VirtualInfrastructureManager_win64\ win6\AssetImplConfig.xml
   • assetImpl_coServerMaxThreadCount = 100
   • assetImpl_coServerFragmentSize = 16384
   • assetImpl_coServerReceiveWaitTime = 30
   • assetImpl_coServerResponseTimeoutTime = 3600
   • assetImpl_cacheTimeout = 3600
2. Change the following parameters listed in the vmware.properties file located at:
   %installationDirectory%\BMC Software\BladeLogic\RSCD\daal\Implementation\BMC_VMware_VirtualInfrastructureManager_win64\win64

• 
  
  • com.bladelogic.vmware.task.timeout=10800000 (milliseconds)                
  • clone-timeout=10800000 (milliseconds)                              
  • customization-timeout=10800000 (milliseconds)
  • customization-timeout-win2k8=10800000 (milliseconds)

Oracle database tuning

BMC Compact RDS provides the flexibility of migrating the database from Oracle Express to the Enterprise Oracle database. While tuning the Oracle database instance, use the following checklist:

1. Make sure that alert.log is clean.
2. Use an appropriately sized SGA.
3. Set cursor_sharing to FORCE.
4. Increase Processes to 500.
5. Run gather stats.

Less-than-optimal VM cloning performance at ESX server or VC server requires additional tuning

The following set of parameters was configured to achieve optimal performance during benchmarking tests. You might need to further tune these parameters' values to achieve optimal performance under different workload and infrastructure conditions. 

Virtual machine cloning time on the VC depends on the underlying hardware, network connectivity, and Configurations Max parameters from VMware.

For environments with optimal hypervisor and storage performance, BMC recommends that you use the default settings. But for environments with slower hypervisor or storage, change the following parameters in the provider.json file (located at BMC Software\BMCCloudLifeCycleManagement\Platform_Manager\configuration\ provider.json).