Power8 Performance Best Practices

Steps

• Ensure that firmware is current
  Fix Central provides latest updates. Install the latest version of firmware for all Power8 systems.
  See the Power8 System Firmware website
  NOTE: Ensure that required HMC level is installed when updating firmware.
   
• Memory DIMMs
  For optimal performance on workloads that are memory bandwidth sensitive follow these recommendations.
  Follow proper memory plug-in rules in the Technical Overview and Introduction links for the following systems:
  S812L and S822L
  S822
  S824L
  E850
  E870 and E880
  
  S814/S824/S822/S822L:
  • Assign minimum 4 DIMMs per socket.
  • 2 DIMMs behind each Power8 chip with all same size DIMMs.
  
  E870/E880:
  • Assign a minimum of 4 DIMMs of same size per socket.


• Ensure OS level is current
  Fix Central provides the latest updates for AIX, IBM i, VIOS, Linux, HMC and firmware. In addition to that, the FLRT tool provides the recommended levels for each hardware model. Use these tools to maintain your system levels.
   
• 40GbE adapter
  It is recommended that this adapter (FC EC3A & EC3B) be installed on the internal CEC slots.
  RHEL7 and onwards: For network bandwidth sensitive workloads, we recommend you increase the receive queue size from 1024 to 8192.
   
• Sizing a system
  • Use the Enhanced rPerf values in the System Performance Reports to correctly size Power8 systems.
  Scale down both entitled capacity and virtual CPUs of a partition in accordance with relative rPerf value, for similar performance.
  • Use Workload Estimator (WLE) for sizing LPARs for CPU consumption as it provides better sizing results.
   
• Right-size your Shared LPARs
  • Assign entitled capacity (EC column in vmstat) to sustain peak utilization for LPARs with critical SLA requirement.
  • Assign EC to average utilization and number of virtual CPUs to peak utilization (physical core consumption) for LPARs with non-critical SLA.
  • Ensure the average LPAR utilization is equal or less than 75% of the entitled capacity.
   
• Partition placement
  • Use Dynamic Partition Optimizer (DPO)  to optimize placement after performing DLPAR operations (memory or CPU) or LPM operations.
  • Ensure that LPAR affinity score is above 90, for best performance.
  NOTE: Additional memory or CPU resources may be required to achieve a score of 90 or above.
   
• Websphere
  Install appropriate levels. See the following document
   
• Java
  • IBM Java 7.1 SR1 is the minimum level to exploit Power8. It is recommend to be at least at IBM Java 8 SR1 FP10 or later.
  7.0 and below defaults to ppc-common instructions set.
  • Use of 64k size pages increases application  performance.
  • If the application uses encryption, try H/W acceleration for encryption by setting Dcom.ibm.crypto.provider.doAESInHardware=true
  •  IBM Semeru Runtimes Downloads are available here
   
• IBM i
  Ensure Technology Updates are current.
  These can be found here.
   
• AIX Tunables / VIOS Tunables
  • Tuning a VIOS is not recommended unless directed by VIOS/AIX support.
  • Restricted tunables should not be modified (unless directed by AIX/VIOS development).
  • Tunables should not be migrated across hardware or AIX levels.
  • If tunables have been modified from default, they should be re-evaluated prior to moving to a different processor platform.
   
• AIX CPU utilization
  The system is optimized for best raw throughput at higher CPU usage. If you need to reduce CPU usage, use the schedo  tunable vpm_throughput_mode to tune the workload and evaluate the benefits of raw throughput vs. CPU usage.
   
• Live Partition Mobility (LPM)
  • Migration of LPARs (in VIO only mode) to Power8 system (Power6/Power7 compat mode):
  ➢ Manual Migration (via backup/restore) should work as is.
  ➢ LPM operations using FC NPIV may need Live Partition Mobility  (LPM) ifix based on migrating LPAR’s AIX levels. Please use the Fix Level Recommendation Tool (FLRT) tool to see the requirements (OS, VIOS, HMC and F/W) for a given migration.
   
• VIOS configuration
  • Assign total entitlement of all VIOS partitions to be 10-15% of cores in shared pool. Refer to the PowerVM Best Practices for additional recommendations.
  • If configured as a shared partition, assign a CPU ratio of 2:1 (vCPU:ec) to the VIOS.
  • Assign uncapped mode and set variable weight capacity of VIOS partition higher than all client LPARs serviced by VIOS.
  • For performance and flexibility, it is recommended to use IBM i to virtualize internal storage to IBM i. If you must use VIOS, follow the  wiki at the following  link.
  • For vFC, ensure no more than 64 client connections total per physical fcs adapter on the VIOS. Also, ensure no more than 64 storage ports configured per  vFC adapter on the client. These are physical limits:
  practical limits may differ based on workload.
  • For vSCSI disks, ensure the queue_depth for virtual disks is less than or equal the queue_depth of the physical disk in the VIOS.
  • For vSCSI adapters, ensure you configure VTDs based on the following formula:
  Max VTDs = ( 512 -2 ) / (virtual_q_depth + 3)
  • Only enable the largesend attribute on the SEA (physical adapter backing the SEA) if all LPARs serviced by the VIOS are AIX partitions.
   
• Virtual Ethernet adapters on AIX
  • Increase the virtual Ethernet (vETH) device driver buffers if the partition is dropping packets on the virtual interface even when running with entitled CPU  capacity. e.g.
  # chdev –l ent# -a max_buf_xxx=NNNN
  • Set largesend on vETH adapter to improve performance (AIX):
  # chdev –l en# -a mtu_bypass=on (or) ifconfig en# largesend