Computation details for VMware metrics


This topic provides computation information for VMware metrics and indicators. 

All metrics used in this computation are based on the 95th percentile values of hourly samples of these metrics over the last 30 days. The top 5% spike is not considered. 

VMware - Memory Pressure Index

Memory Utilization for clusters, hosts, and VMs reported by VMware, provide only one aspect of memory contention. It does not take into account many other factors that affect memory demand. In addition, memory utilization for clusters, hosts, and VMs is based on Consumed Memory which gives an inflated picture of memory usage. In addition to Consumed Memory, VMware reports Active Memory that provides a deflated picture of memory usage.

Memory Pressure Index (MPI) takes into account consumed memory, active memory, and other metrics like ballooning and swapping and provides a realistic picture of the memory pressure in VMware Clusters, Hosts, and VMs (Windows VMs only). 

MPI is used to compute Composite Memory Utilization and Composite Memory Consumption. Composite Memory Utilization and Composite Memory Consumption are alternate values for Memory Utilization and Consumed Memory computed using MPI.

The following tables describe the computation and other important details for Memory Pressure Index (MPI):

MPI calculation for Clusters and Hosts

The following points about MPI for clusters and hosts should be noted:

  • MPI Value: MPI will have a value in the range of 0 – 100. The higher the number, the higher the pressure.
  • Minimum value: Minimum value of memory pressure = Active Memory Utilization
  • Penalty points: There are penalty points for Consumed/Active RatioBallooningOver-commitment, and Swapping scenarios, which are described in the following table.

MPI
Calculation

Penalty

Threshold

Penalty

Computation

Min

Max

Condition

Minimum
Penalty
multiplier
factor


C/A RATIO >=


Part 1

Consumed/Active Ratio
(C/A Ratio)

Configurable

Default: 6

5

10

2 x Threshold

2

1 x Threshold

1


BALLOONING >=


Part 2

Ballooning

Configurable
Default: 1 GB 

5

15

3 x Threshold

3

2 x Threshold

2

1 x Threshold

1


OVER-COMMITMENT >=


Part 3

Over-commitment

Configurable

Default: 1.2 

5

15

Threshold + 0.4

3

Threshold + 0.2

2

Threshold

1


SWAP_RATE >=


Part 4

Swap Rate

Configurable

Default: 64 MB/sec

5

30

6 x Threshold

6

5 x Threshold

5

4 x Threshold

4

3 x Threshold

3

2 x Threshold

2

1 x Threshold

1

Estimated MPI

If Part 2 = 0:
Estimated MPI = Active Memory Utilization + Part 1 + Part 3 + Part 4

Else:
Estimated MPI = Active Memory Utilization + Part 2 + Part 3 + Part 4


Actual MPI

Actual MPI = MIN (Consumed Memory Utilization, Estimated MPI)
 

MPI Calculation for VMs (Windows only)

The following points about MPI for VMs should be noted:

  • MPI value: MPI will have a value in the range of 0 – 100. The higher the number, the higher the pressure.
  • Minimum value: Minimum value of memory pressure = Memory Utilization/2.
  • Penalty points: The penalty is calculated based on the value of the memory swap rate.

Metrics/Parameters Used in MPI Calculation

Metrics/Parameters
used in MPI calculation

Threshold

Penalty

Computation

Min

Max

Memory Swap Out Rate

MEM_SWAP_OUT_BYTE_RATE

32 Mbytes/sec

10 

 50

Swap Rate threshold exceed count

1



count(MEM_SWAP_OUT_BYTE_RATE > threshold (MEM_SWAP_OUT_BYTE_RATE))





Condition

Penalty for Swap Rate





Swap Rate threshold_exceed_count > 5 * Minimum swap rate exceed count 

5 * MINIMUM_PENALTY

Swap Rate threshold_exceed_count > 4 * Minimum swap rate exceed count 

4 * MINIMUM_PENALTY

Swap Rate threshold_exceed_count > 3 * Minimum swap rate exceed count 

3 * MINIMUM_PENALTY

Swap Rate threshold_exceed_count > 2 * Minimum swap rate exceed count 

2 * MINIMUM_PENALTY

Swap Rate threshold_exceed_count > 1 * Minimum swap rate exceed count 

1 * MINIMUM_PENALTY

Memory Pressure Index = Memory Utilization/2 + Penalty for Swap Rate

Composite Memory computation

Composite Memory is stored as an Indicator and is calculated through the following formula:

Composite Memory Consumption

Min ( Consumed Memory, (Active Memory + ((Consumed Memory – Active Memory)
x Memory Pressure Index /100)))

Composite Memory Utilization

Min ( Consumed Memory Util, (Active Memory Util +
((Consumed Memory Util – Active Memory Util) x Memory Pressure Index /100)))

VMware - CPU Pressure Index

CPU Utilization of the cluster or host, reported by VMware, provides the overall CPU consumption of the cluster or host but will not reflect any problems with capacity or performance of individual VMs.

The CPU Pressure Index (CPI) uses the CPU Ready of the VMs in addition to the CPU Utilization to present a better picture of CPU contention in the cluster or host. It is an index between 0 and 100. The higher the value, the higher the pressure. This metric is calculated for VMware hosts and clusters.

The following points about CPI should be noted:

  • CPI Value: CPU Pressure has a value in the range of 0 – 100. The higher the number, the higher the pressure.
  • Minimum value: Minimum value of CPU pressure = CPU Utilization. If the CPI value is higher than the CPU Utilization, it indicates that some of the VMs are not getting enough CPUs. Review the CPU Starved VMs tab on the Watchlist page in the vSphere Virtual Machines view to check if the VMs running in the cluster have high CPU time. Accordingly, you need to allocate more resources to such VMs. 
  • Penalty Points: There are Penalty Points for Cluster CPU Ready Time and Cluster CPU Ready Exceed Count scenarios, which are described in the following table with computation and other details for CPU Pressure Index (CPI). 

MPI
Calculation

Penalty

Threshold

Penalty

Computation




Min

Max

Condition

Minimum

Penalty

multiplier

factor


CPU_READY >=


Part 1

Cluster CPU Ready:

Cluster CPU Ready =
Average of CPU Ready
of VMs that exceed CPU
Ready threshold


Configurable

Default: 10%


5

20

4 x Threshold

4

3 x Threshold

3

1 x Threshold

1


COUNT >=


Part 2

Cluster CPU Ready Exceed Count:

Cluster CPU Ready Exceed Count =
Number of VMs that
exceed threshold

Configurable
Default: 10

5

20

4 x Threshold

4

3 x Threshold

3

2 x Threshold

2

1 x Threshold

1

Combined CPI

If CPU_Utilization > CPU_Util_thresh, then:

CPU Pressure Index = CPU_Utilization

Else:

CPU Pressure Index = CPU Utilization + MAX (Part 1, Part 2)

VMware - Storage Pressure Index

An index between 0 and 100 is calculated using Storage utilization and Datastore I/O latency. The higher the value, the higher the pressure. This metric is calculated for vSphere hosts and clusters.

The following points about SPI should be noted:

  • SPI Value: The Storage Pressure Index will be between 0 and 100. The higher the number, the higher the pressure.
  • Minimum value: Minimum value of Storage pressure = Storage Utilization
  • Penalty Points: There are Penalty Points for High I/O Latency Time scenarios, which are described in the following table with computation and other important details for Storage Pressure Index (SPI):

SPI
Calculation

Penalty

Threshold

Penalty

Computation




Min

Max

Condition

Minimum

Penalty

multiplier

factor


LATENCY >=


Part 1

Cluster I/O Latency:

Cluster I/O Latency =
Average of I/O Latency of
datastores that exceed I/O latency threshold


Configurable

Default: 30 msec

5


20


4 x Threshold

4

3 x Threshold

3

2 x Threshold

2

1 x Threshold

1


COUNT >=


Part 2

Cluster I/O Latency  Exceed Count:

Cluster IO Latency Exceed Count =
Number of Datastores that exceed threshold


Configurable
Default: 10


5

20

4 x Threshold

4

3 x Threshold

3

2 x Threshold

2

1 x Threshold

1

Combined SPI

If Storage_Utilization > Storage_Util_thresh, then:

Storage Pressure Index = Storage_Utilization

Else:

Storage Pressure Index = Storage Utilization + MAX (Part 1, Part 2)

 

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