Observability Costs & Signals: Log Volumes, Metric Cardinality, Trace Sampling, Retention Tiers & Alert Noise

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Turn events per second and bytes per event into GB per day for eight log streams, add them up per service with SUMIFS and work out the monthly ingest that a log bill is based on.
• Count the possible time series of a metric with PRODUCT, see what one label with thousands of values does to it, and price the series that really occur.
• Compare a head-sampling rule with a tail rule that keeps every error, count the error traces each one keeps or loses, and size the stored span volume.
• Turn a retention policy into GB stored per tier, price each line with INDEX/MATCH from a tier table and see what keeping audit logs searchable for longer really costs.
• Price five teams on three fictional plans, use MAX to bill only the GB above an allowance, and let MIN and a lookup name the cheapest plan for each team.
• Price three savings measures, apply them in the right order so no GB is saved twice, and express the total as a share of today's bill next to what each measure costs in visibility.
• Measure a two-week alert log: non-actionable pages per week with COUNTIFS, the share of on-call minutes they waste, repeats on the same day, and you colour the noise with a formula rule.
• Build the monthly observability report: cost per service, cost per 1,000 requests, the variance against budget with an Over/Within status, and a stacked column chart of where the money goes.