Solving performance problems

Ops Center Analyzer User Guide

Version
11.0.x
Audience
anonymous
Part Number
MK-99ANA002-06

The common performance problems and the possible solutions are described as follows. The possible causes and solutions are intended to provide guidance, and might not satisfy your business process performance requirements.

The following table lists the commonly observed storage-related problems and possible solutions.

Bottleneck metrics Root cause and possible solutions
Parity Group utilization
  • Root cause:
    The usage rate of the Parity Group increases because of the following possible causes:
    • Some volumes might be under heavy load.
    • Volumes (logical resources) might belong to the same Parity Group (physical resource) which might cause resource contention issues in the shared infrastructure.
  • Possible solutions:
    • Consider moving some volumes to another Parity Group with a lower usage rate or higher performance.
    • Consider increasing the number of drives (by concatenating Parity Groups).
    • To manage a Parity Group that is part of a pool, consider adding another Parity Group to the pool.
MPB utilization
  • Root cause:

    The usage rate of the MP Blade (average usage rate of the MP cores in the MP Blade) increases because of an increased load. Too many busy resources such as internal volumes, external volumes, or journal groups accessing the same MP Blade might cause performance degradation.

  • Possible solutions:
    • Consider allocating the busy resources (internal volumes, external volumes, or journal groups) to another MP Blade (changing the MP Blade ownership).

    • Limit I/O to the volumes using I/O controls.

    • Increase the size of cache memory allocated to the MP Blade.

Port utilization
  • Root cause:

    The usage rate of the port (amount of data forwarded by the port divided by the amount of data that can be forwarded by the port) increases because of a number of volumes accessing the same port.

  • Possible solutions:

    Consider allocating some volumes (or host groups) to a different port.

    Note: When the connected port is changed, the host might need to be restarted.

Cache utilization
  • Root cause:

    Out of the total cache memory allocated to the CLPR, the percentage occupied by the data waiting to be written to the drive increases. The cache write pending rate increases because of the following possible causes:

    • The usage rate of the drive might be high, delaying write processing to the drive.

    • The usage rate of the processors might be high, delaying write processing to the drive.

    • The capacity of the installed cache memory might be insufficient.

  • Possible solutions:
    • Consider allocating some volumes to another cache partition.

    • Consider increasing the cache memory.

    • Limit I/O to the volumes using I/O controls.

    • Consider allocating the busy resources (internal volumes, external volumes, or journal groups) to another MP Blade (changing the MP Blade ownership).

The following table lists the commonly observed server-related problems and possible solutions.

Bottleneck metrics Root cause and possible solutions
CPU utilization
  • Root cause:

    CPU bottlenecks occur when several VMs run on the same physical machine, and end up sharing the same CPU. If the VMs (logical resources) share the same CPU (physical resource) and if one of the VMs utilizes the CPU more than the others in the shared infrastructure, the total efficiency of the CPU is degraded and the CPU utilization rate increases. The CPU could become saturated with requests because of resource contention issues.

  • Possible solutions:

    Consider moving the VMs to another server.

Memory utilization
  • Root cause:

    Memory bottlenecks occur when several VMs (logical resources) share the available memory (physical resources) which might result in the performance degradation of the physical memory.

  • Possible solutions:

    Consider allocating additional physical memory, or moving the VMs to another server.