Sizing formulas

Edit online

This section provides a summary of the sizing formulas. You can use these formulas as a guideline to size your deployments.

The formulas are designed to estimate the IFL demand, which is the minimum number of physical IFLs required for a given deployment. Keep in mind that performance varies with configuration and other factors. Therefore, results that are obtained in other operating environments might vary. We recommend that you run your own experiments with a configuration applicable to your environment to measure the IFL demand.

IFL demand

Each formula uses the IFL demand as measure, which expresses the number of additional IFL cores that are used to operate the component. In this regard, it is the difference between the used IFL cores during operation and the used IFL cores of a specific baseline.

To compute the used IFL cores, the real utilization formula for SMT-2 enabled systems is used. The assumption is an SMT speedup of 1.3. This formula uses the hyptop metrics core utilization $u_{c}$ , thread utilization $u_{t}$ , and the management utilization $u_{m}$ to compute the real IFL utilization:

D_{u t i l} = \frac{2 u_{c} - u_{t}}{1.3} + (u_{t} - u_{c}) + u_{m}

The used IFL core demand is calculated by dividing it by 100 to express the utilization percentage as the number of used cores:

D_{I F L} = \frac{D_{u t i l}}{100}

Note: The calculation based on formulas is abstract. Therefore, in each section, an illustrating example based on a bare metal deployment is provided.

RHOCP cluster sizing

Estimates how many IFLs are required for a cluster with a specific number of RHOCP nodes:

D_{I F L} = N * C_{n o d e} + D_{m g m t}

where:

$N_{}$: Number of compute nodes
$C_{n o d e}$: IFL demand per compute node
$D_{m g m t}$: IFL demand for control plane node

Red Hat OpenShift Virtualization (RHOCPV) sizing

Estimates how many IFLs are required to have a specific number of idling VMs in a running and ready state:

D_{I F L} = N * C_{v m} + D_{o p e r a t o r}

where:

$N_{}$: Number of idling VMs
$C_{v m}$: IFL demand per idling VM
$D_{o p e r a t o r}$: IFL demand for the Red Hat® OpenShift® Virtualization Operator

RHOCP Operator sizing

Estimates how many IFLs are required to have a specific number of VMs in running and ready state:

D_{I F L} = D_{o p e r a t o r}

where:

$D_{o p e r a t o r}$: IFL demand for Operator deployment

RHOCP sizing for infrastructure add-ons

Estimates the IFL demand for a workload with a specific number of transactions-per-second using a specific service infrastructure add-on:

D_{I F L} = N * C_{t p s} + D_{o p e r a t o r}

where:

$N_{}$: Number of transactions per second
$C_{t p s}$: IFL demand per transaction per second
$D_{o p e r a t o r}$: IFL demand for RHOCP infrastructure Operator