Setting up Linux on IBM Z systems


The resources for operating BMC AMI Ops Console Management are primarily a function of the number of host consoles and the volume of messages sent to those consoles by the host LPARs.

Mainframes are typically highly utilized with constant message traffic so BMC AMI Ops Console Management will correspondingly require significant continuous use of CPU and memory with occasional extra CPU demands.

Related topic

Planning-and-installing

The minimum requirement for a VM-managed zLinux image is:

  • 1.5 Gb of physical memory
  • 24 MSUs of CPU with up to 72 MSUs available on demand for short (less than 1 minute) periods
  • 10 Gb of DASD

This would be sufficient for a small site with 20 active consoles and 3 operators monitoring those consoles using an out-of-the-box BMC AMI Ops Console Management installation, whereas an active console is defined as one receiving 25,000 messages an hour (approximately 7 messages a second).

For each additional 10 consoles, add to the above:

  • 256 Mb of physical memory
  • 8 MSUs of CPU (base, add 12 MSUs to the peak requirement)

The cost for additional operators (client connections) varies depending on how many consoles are monitored by each operator; a rough estimate is an additional 128 Mb of physical memory and an MSU per operator.

Use of the BMC AMI Ops Console Management rules language to perform custom message processing can significantly increase CPU requirements; any increase in the size of the BMC AMI Ops Console Management message cache size from the default of 10,000 will significantly increase both memory and CPU usage.

Some additional notes:

  • Memory—the allocated physical memory must be enough to prevent both zLinux and VM paging activity. A number of settings changes to the Linux memory system defaults are highly recommended in order to optimize zLinux for the BMC AMI Ops Console Management workload.
  • CPU—the BMC AMI Ops Console Management zLinux image must be marked as quick dispatch ('cp set quickdsp') and have sufficient minimum and maximum capacity assigned to meet demands calculated above. If the image is hosted on a VM LPAR spanning multiple physical IFLs then it will benefit from being assigned virtual processors equal to the number of physical processors.
  • DASD—the 10 Gb specified really is a bare minimum, 20 Gb is recommended.

The scaling factors (cost for additional consoles and/or operators) above are rough estimates: the example zLinux platform is a business class z10 IFL rated at 83 MSUs (673 MIPs). The zLinux image was assigned 2 Gb of physical memory and CPU capped at a maximum of 50% (ie. ~40 MSUs) running 20 console & 3 operator load test.

The IBM relative performance ratings (MSUs) for a single IFL on various Z platforms:

  • z10 Business Class (BC10) = 83 MSUs
  • z10 Enterprise Class (z10 EC) = 115 MSUs
  • z114 (Business Class) = 98 MSUs
  • z196 (Enterprise Class) = 150 MSUs
  • z12 (Enterprise Class) = 188 MSUs


 

Tip: For faster searching, add an asterisk to the end of your partial query. Example: cert*