Lab: Design Super Metrics & Custom Groups for Cost Optimization

VCF 9.0Advancedvcap-advanced⏱ 120 min

VCF Operations (formerly Aria Operations) — super metrics formula language, custom group dynamic membership, chargeback/showback cost models

Objectives

Create super metrics using formula language with cross-object aggregation
Build custom groups with dynamic membership rules for business-context organization
Design a cost allocation model using super metrics for per-VM and per-group chargeback
Identify underutilized VMs using efficiency super metrics and dynamic group membership
Export super metric definitions for version control and environment migration

Prerequisites

VCF 9.0 with VCF Operations deployed and collecting metrics from vCenter for at least 7 days (required for meaningful efficiency analysis). Multiple VMs across clusters with varying utilization levels.

Required skills:

VCF Operations UI navigation
Basic metrics concepts (CPU usage, memory consumed)
Cost allocation principles

Lab Environment

VCF Operations cluster connected to vCenter managing 2+ clusters with 20+ VMs of varying sizes and utilization patterns. Some VMs should be intentionally oversized or idle to demonstrate rightsizing identification.

Tasks

Task 1 Super Metrics & Custom Groups for Cost Optimization

Super metric design requires understanding the VCF Operations object model — formulas reference specific adapter types, resource kinds, and metric keys. Use 'this' references for metrics on the assigned object, and explicit adapter/resourcekind paths for cross-object aggregation. Always test on a small group first — a formula error applied to 10,000 VMs can overload the analytics engine.

Build a complete cost optimization workflow: create efficiency super metrics that identify waste, build dynamic custom groups that auto-populate with optimization candidates, and design a chargeback model that quantifies savings — demonstrating the full VCF Operations analytics pipeline from raw metrics to business-level cost reporting.

Step 1

Navigate to super metric creation. VCF Operations → Configure → Super Metrics → Add. Name='vm_cpu_efficiency_pct'. Object Type=VirtualMachine. Formula: (${this, metric=cpu|usagemhz|average} / (${this, metric=cpu|corecount_provisioned} * ${this, metric=cpu|cpuMhz})) * 100. This calculates CPU efficiency as: actual MHz consumed / total MHz provisioned × 100. A VM with 4 vCPU @ 2600 MHz (10,400 MHz provisioned) using 1,040 MHz averages 10% efficiency — a clear rightsizing candidate.

Step 2

Create memory efficiency super metric. Add → Name='vm_memory_efficiency_pct'. Object Type=VirtualMachine. Formula: (${this, metric=mem|consumed|average} / ${this, metric=mem|guest|provisioned}) * 100. This shows actual memory consumption vs provisioned — capturing balloon/swap scenarios where a VM appears to use all allocated memory but actually needs far less. A value < 30% sustained over 30 days indicates oversized memory allocation.

Step 3

Create per-VM cost super metric. Add → Name='vm_monthly_cost_usd'. Object Type=VirtualMachine. Formula: (${this, metric=cpu|usagemhz|average} * 0.02) + (${this, metric=mem|consumed|average} / 1048576 * 5.00) + (${this, metric=virtualDisk|totalReadLatency|average} * 0.001). This assigns dollar costs based on actual resource consumption: $0.02/MHz CPU + $5.00/GB memory + $0.001/ms disk latency. Adjust rates to match your organization's infrastructure cost model.

Step 4

Create cluster-level contention super metric. Add → Name='cluster_cpu_contention_score'. Object Type=ClusterComputeResource. Formula: avg(${this, metric=cpu|ready|summation, depth=1}) / avg(${this, metric=cpu|usagemhz|average, depth=1}) * 100. This aggregates CPU ready time across all hosts in the cluster (depth=1 = direct children) and normalizes by utilization. Values > 5% indicate the cluster is over-committed — VMs are waiting for CPU time.

Step 5

Enable super metrics on object types. For each super metric, click the 'Object Types' tab and select the target: VirtualMachine for VM-level metrics, ClusterComputeResource for cluster metrics. Enable 'Monitor' to start data collection. VCF Operations will calculate and store these values at each collection interval (5 minutes). Wait 15-20 minutes for initial data population before proceeding.

Step 6

Create dynamic custom group for underutilized VMs. Configure → Custom Groups → Add. Name='Rightsizing-Candidates-Oversized'. Group Type=VirtualMachine. Membership: Define Membership Criteria → Add Rule: 'Super Metric vm_cpu_efficiency_pct is less than 20' AND 'Super Metric vm_memory_efficiency_pct is less than 30' AND 'Property Summary|Runtime|Power State equals poweredOn' AND 'Property Summary|Runtime|BootTime is more than 30 days ago'. Click Preview Members to verify the dynamic query returns expected VMs. This group auto-updates as VM utilization changes.

Step 7

Create custom group for idle VMs. Add → Name='Decommission-Candidates-Idle'. Membership: 'Super Metric vm_cpu_efficiency_pct is less than 2' AND 'Metric net|usage|average is less than 10 KBps' AND powered on for > 30 days. These VMs are consuming resources but doing essentially nothing — prime candidates for decommissioning. Create a third group: 'Cost-Centers-Engineering' with membership based on VM tag or folder path matching the engineering team's resources.

Step 8

Build chargeback cost report. Navigate → Dashboards → Create Dashboard. Name='Cost-Optimization-Dashboard'. Add widgets: (a) Scoreboard: 'Total Monthly Cost' showing sum of vm_monthly_cost_usd across all VMs; (b) Top-N: 'Top 10 Most Expensive VMs' ranked by vm_monthly_cost_usd; (c) Scoreboard: 'Rightsizing Savings Potential' showing count of Rightsizing-Candidates group members × average cost difference; (d) List: Members of 'Decommission-Candidates-Idle' with current cost. Use widget interactions so clicking a VM in the Top-N updates a metric chart showing its efficiency trend.

Step 9

Create a scheduled report for management. Configure → Reports → Add Report Template. Name='Monthly-Cost-Optimization'. Add views: (a) List view of Rightsizing-Candidates with columns: VM Name, Current CPU, Recommended CPU, Current Memory, Recommended Memory, Monthly Savings; (b) Summary view of Cost-Centers-Engineering showing total cost, VM count, average efficiency. Schedule: Monthly, first Monday, email to stakeholder distribution list. Format: PDF.

Step 10

Export super metric definitions for version control. Configure → Super Metrics → select all → Export. This downloads an XML file containing all formula definitions, object type bindings, and descriptions. Store in Git for version control. To migrate to another VCF Operations instance: Import the XML file on the target. This enables dashboard-as-code and metrics-as-code practices — critical for multi-environment consistency.

Validation Gate

Check: Super metrics calculating, custom groups auto-populating, and cost dashboard operational

Expected: 4 super metrics computing values for all VMs and clusters, 3 dynamic custom groups with correct membership based on efficiency thresholds, cost dashboard showing per-VM costs and rightsizing savings potential, scheduled report configured for monthly delivery

Common Errors

Super metric shows 'No Data' after creation

Fix: Check: (1) Object Type assignment — the metric must be enabled on the correct resource kind; (2) formula syntax — use the built-in formula validator; (3) wait 15-20 minutes for initial calculation; (4) the referenced metrics must exist on the target objects (e.g., cpu|usagemhz|average requires the vCenter adapter to be collecting).

Custom group shows 0 members despite matching VMs

Fix: Dynamic membership criteria may reference metrics or super metrics that haven't been calculated yet. Verify: (1) super metric has data (check in a metric chart); (2) criteria operators are correct (less than vs greater than); (3) property names match exactly (case-sensitive). Use 'Preview Members' to debug the query.

Cost calculation formula gives unreasonable values

Fix: Unit mismatch in the formula. VCF Operations metrics use specific units: cpu|usagemhz is in MHz, mem|consumed is in KB (not GB). Verify unit conversion factors in your formula: divide memory by 1048576 to convert KB to GB.

Super metric performance degradation with many objects

Fix: Formulas using depth=-1 (recursive) on large hierarchies can be expensive. Limit to depth=1 where possible. Also: avoid creating more than 50 super metrics per object type — each one adds processing overhead at every collection cycle.

Final Validation

Cost optimization workflow operational with super metrics, custom groups, and automated reporting

✓ CPU efficiency super metric → Calculating percentage values for all VMs (0-100 range)

✓ Memory efficiency super metric → Calculating percentage values for all VMs

✓ Per-VM cost super metric → Dollar values calculated for all VMs

✓ Rightsizing candidates group → Auto-populated with VMs matching efficiency criteria

✓ Cost dashboard → Displaying Top-N costs, savings potential, and drill-down interactions

✓ Scheduled report → Configured for monthly delivery with correct views and distribution

Cleanup / Restore

• Delete scheduled report

• Delete custom groups

• Delete dashboard

• Delete super metrics (in reverse order — dashboard first, then groups, then metrics)

Design Reflection (VCDX)

Cost optimization through data-driven rightsizing is a compelling VCDX story. It demonstrates: (1) Technical depth — understanding VCF Operations formula language and object model; (2) Business alignment — translating technical metrics into financial impact; (3) Operational maturity — automated reporting and dynamic grouping reduce manual effort. In defense, be prepared to justify your cost model rates and explain how dynamic thresholds avoid false positives in rightsizing recommendations.

Requirements

Identify 20%+ cost reduction opportunity through rightsizing and decommissioning
Automated monthly reporting to finance/management
Dynamic candidate identification without manual VM-by-VM review

Constraints

Super metric performance degrades beyond 50 formulas per object type
Rightsizing requires 30+ days of historical data for reliable recommendations
Cost model rates must be calibrated to actual infrastructure costs (hardware, power, licensing)

Assumptions

VMs have been running with stable workloads for 30+ days
Cost model rates approved by finance team
Rightsizing changes are applied during maintenance windows (not live)

Risks

Aggressive rightsizing (based on average instead of peak) causes performance degradation during burst periods
Idle VM decommission without business validation removes disaster recovery or seasonal workloads
Super metric formula errors produce incorrect cost data — misleading management decisions

Self-Assessment Discussion Prompts

How do you calibrate chargeback rates to reflect actual infrastructure costs?
When should you use conservative (peak-based) vs aggressive (average-based) rightsizing?
How do you handle seasonal workloads that appear idle for 10 months but are critical during peak periods?
What governance process ensures rightsizing recommendations are validated before execution?

Extensions

Build a super metric that calculates 'cost waste' = (provisioned cost - consumed cost) for each VM

Create a Slack webhook notification when the Rightsizing-Candidates group grows beyond a threshold

Implement a rightsizing automation pipeline using VCF Operations API + vSphere API

Design a multi-tenant chargeback model with per-department custom groups and separate cost rates

⚠ Known Pitfalls (from Community KB)

Using depth=-1 in super metrics on large hierarchies — recursive aggregation across 10,000+ objects degrades collection performance

Not validating cost model rates with finance — technical cost estimates that don't match actual billing create trust issues

Rightsizing based on 7-day data instead of 30+ days — misses weekly peaks and monthly batch jobs

Forgetting to exclude template VMs and infrastructure VMs from rightsizing groups — these have intentionally low utilization

References

VCF Operations Administration Guide — Super Metrics: techdocs.broadcom.com
VCF Operations API Reference — /supermetrics endpoint: techdocs.broadcom.com

Mark complete