# Example 734 — The budget alert was on spend and the commitment was on capacity `the_budget_alert_was_on_spend_and_the_commitment_was_on_capacity.eml` - Budget alerts are configured at two thresholds, tested with a real charge, and routed to a person who acknowledges them. What they can move with is computed below. ## EML ```eml # Self-authored for the EML case corpus (no external origin). Budget alerts are # configured at two thresholds, tested with a real charge, and routed to a # person who acknowledges them. What they can move with is computed below. # # The alerting is set up properly. There are two thresholds rather than one, so # there is a warning before the wall; the configuration was tested by pushing a # synthetic charge through and watching the alert arrive, rather than trusting # the console; it is routed to a named person rather than a distribution list; # and that person acknowledges. Nobody here was careless. # # The alert fires on SPEND. The account is on a three-year committed-use # agreement, so the invoice is the commitment whether the capacity is used or # not, and spend is constant by construction. # # Utilisation went from eighty-two percent to thirty-one after a migration. 2 => alert_thresholds_configured 1 => synthetic_charge_tests 1 => named_people_who_acknowledge 36 => commitment_months 420000 => monthly_commitment_dollars 82 => utilisation_percent_before 31 => utilisation_percent_after 100 => a_whole_percent 7 => months_since_the_migration 0 => alerts_that_fired 0 => alerts_configured_on_utilisation 0 => month_over_month_spend_variance_dollars a_whole_percent - utilisation_percent_after => unused_percent_after utilisation_percent_before - utilisation_percent_after => utilisation_percentage_points_lost int(monthly_commitment_dollars * unused_percent_after / a_whole_percent) => monthly_dollars_for_unused_capacity monthly_dollars_for_unused_capacity * months_since_the_migration => dollars_for_unused_capacity_since "alert thresholds configured : " + str(alert_thresholds_configured) ^0 "synthetic charge tests : " + str(synthetic_charge_tests) ^0 "named people who acknowledge : " + str(named_people_who_acknowledge) ^0 "alerts that fired : " + str(alerts_that_fired) ^0 "" ^0 "commitment months : " + str(commitment_months) ^0 "monthly commitment, dollars : " + str(monthly_commitment_dollars) ^0 "month over month spend variance : " + str(month_over_month_spend_variance_dollars) ^0 "" ^0 "utilisation before, percent : " + str(utilisation_percent_before) ^0 "utilisation after, percent : " + str(utilisation_percent_after) ^0 " points lost : " + str(utilisation_percentage_points_lost) ^0 " unused after, percent : " + str(unused_percent_after) ^0 "" ^0 "monthly dollars for unused capacity : " + str(monthly_dollars_for_unused_capacity) ^0 "months since the migration : " + str(months_since_the_migration) ^0 " dollars since : " + str(dollars_for_unused_capacity_since) ^0 "alerts configured on utilisation : " + str(alerts_configured_on_utilisation) ^0 "" ^0 # ---- what the alerting verified ---- "the budget alerts" ^0 " thresholds : " + str(alert_thresholds_configured) + ", so there is a warning before the" ^0 " wall rather than only a wall" ^0 " tested by : pushing a synthetic charge and watching it" ^0 " arrive, not by reading the console" ^0 " routed to : a named person, not a list" ^0 " acknowledged : yes" ^0 " verdict : CONFIGURED AND WORKING" ^0 "" ^0 " testing the alert with a real charge is the step that" ^0 " separates this from a checkbox, and it was done" ^0 "" ^0 # ---- what the alert watches ---- "the monitored quantity" ^0 " what fires the alert : spend crossing a threshold" ^0 " what determines spend : the commitment, for " + str(commitment_months) ^0 " months" ^0 " what changes when capacity goes unused : nothing on the" ^0 " invoice" ^0 " month over month variance : " + str(month_over_month_spend_variance_dollars) ^0 " so the number the alert watches : cannot move" ^0 "" ^0 " the alert is correctly wired to a quantity that this" ^0 " contract holds constant" ^0 "" ^0 # ---- the commitment was a good decision ---- # The three-year agreement was taken deliberately and it saved real money # against on-demand pricing. It is not the mistake here. It is the reason the # instrument stopped being able to move. "the commitment" ^0 " taken deliberately : yes, against on-demand pricing" ^0 " did it save money at the time : yes" ^0 " is it the mistake : no" ^0 " what it does to spend : fixes it for " + str(commitment_months) + " months" ^0 " what it does to the alert : leaves it correctly wired" ^0 " to a constant" ^0 "" ^0 # ---- what the migration changed ---- # The migration was also right: the workload moved to a service that suits it # better and runs faster. It reduced the capacity needed and could not reduce # the capacity bought. "the migration" ^0 " moved the workload to : something that suits it better" ^0 " was that right : yes, on its own terms" ^0 " capacity needed after : lower" ^0 " capacity bought after : the same" ^0 " utilisation, before and after : " + str(utilisation_percent_before) + " and " + str(utilisation_percent_after) ^0 " what appeared on the invoice : nothing" ^0 "" ^0 # ---- the number that would have moved ---- "utilisation" ^0 " is it measured : yes, the provider reports it" ^0 " is it on a dashboard : yes" ^0 " alerts configured on it : " + str(alerts_configured_on_utilisation) ^0 " what it would have shown : " + str(utilisation_percentage_points_lost) + " points, in one month" ^0 " what nobody was watching for : a number going down" ^0 " dollars behind those points, so far : " + str(dollars_for_unused_capacity_since) ^0 "" ^0 # ---- null control ---- # The same alerting, with a threshold on utilisation as well as on spend, and # the commitment renewal gated on the utilisation trend. 2 => nc_alerts_configured_on_utilisation 1 => nc_alerts_that_would_have_fired "null control - a threshold on utilisation too" ^0 " spend thresholds : " + str(alert_thresholds_configured) + ", unchanged" ^0 " utilisation thresholds : " + str(nc_alerts_configured_on_utilisation) ^0 " alerts that fire in the migration month : " + str(nc_alerts_that_would_have_fired) ^0 " the alerting did not get better; it acquired a threshold" ^0 " on a quantity this contract lets move" ^0 "" ^0 # ---- the rule ---- "what a working budget alert guarantees" ^0 " spend crossing a threshold is noticed : exactly, tested" ^0 " with a real charge and acknowledged by a person" ^0 " waste is noticed : not addressed; the alert is" ^0 " denominated in spend, and this contract decouples" ^0 " spend from consumption" ^0 "" ^0 "an alert can only report on a quantity that varies; a" ^0 "contract that fixes the monitored quantity does not break" ^0 "the alert, it removes the alert's subject, and a silent" ^0 "alert is what correct configuration looks like" ^0 "" ^0 "The alerting is done properly: " + str(alert_thresholds_configured) + " thresholds rather than one, tested by pushing a" ^0 "real charge through, routed to " + str(named_people_who_acknowledge) + " named person who acknowledges. It fires on" ^0 "spend, which a " + str(commitment_months) + "-month commitment holds at " + str(month_over_month_spend_variance_dollars) + " variance, so utilisation" ^0 "falling from " + str(utilisation_percent_before) + " to " + str(utilisation_percent_after) + " percent produced " + str(alerts_that_fired) + " alerts while " + str(monthly_dollars_for_unused_capacity) ^0 "dollars a month bought nothing - " + str(dollars_for_unused_capacity_since) + " over " + str(months_since_the_migration) + " months." ^0 ``` ## Python (deterministic transpilation) ```python alert_thresholds_configured = 2 synthetic_charge_tests = 1 named_people_who_acknowledge = 1 commitment_months = 36 monthly_commitment_dollars = 420000 utilisation_percent_before = 82 utilisation_percent_after = 31 a_whole_percent = 100 months_since_the_migration = 7 alerts_that_fired = 0 alerts_configured_on_utilisation = 0 month_over_month_spend_variance_dollars = 0 unused_percent_after = a_whole_percent - utilisation_percent_after utilisation_percentage_points_lost = utilisation_percent_before - utilisation_percent_after monthly_dollars_for_unused_capacity = int(monthly_commitment_dollars * unused_percent_after / a_whole_percent) dollars_for_unused_capacity_since = monthly_dollars_for_unused_capacity * months_since_the_migration print("alert thresholds configured : " + str(alert_thresholds_configured)) print("synthetic charge tests : " + str(synthetic_charge_tests)) print("named people who acknowledge : " + str(named_people_who_acknowledge)) print("alerts that fired : " + str(alerts_that_fired)) print("") print("commitment months : " + str(commitment_months)) print("monthly commitment, dollars : " + str(monthly_commitment_dollars)) print("month over month spend variance : " + str(month_over_month_spend_variance_dollars)) print("") print("utilisation before, percent : " + str(utilisation_percent_before)) print("utilisation after, percent : " + str(utilisation_percent_after)) print(" points lost : " + str(utilisation_percentage_points_lost)) print(" unused after, percent : " + str(unused_percent_after)) print("") print("monthly dollars for unused capacity : " + str(monthly_dollars_for_unused_capacity)) print("months since the migration : " + str(months_since_the_migration)) print(" dollars since : " + str(dollars_for_unused_capacity_since)) print("alerts configured on utilisation : " + str(alerts_configured_on_utilisation)) print("") print("the budget alerts") print(" thresholds : " + str(alert_thresholds_configured) + ", so there is a warning before the") print(" wall rather than only a wall") print(" tested by : pushing a synthetic charge and watching it") print(" arrive, not by reading the console") print(" routed to : a named person, not a list") print(" acknowledged : yes") print(" verdict : CONFIGURED AND WORKING") print("") print(" testing the alert with a real charge is the step that") print(" separates this from a checkbox, and it was done") print("") print("the monitored quantity") print(" what fires the alert : spend crossing a threshold") print(" what determines spend : the commitment, for " + str(commitment_months)) print(" months") print(" what changes when capacity goes unused : nothing on the") print(" invoice") print(" month over month variance : " + str(month_over_month_spend_variance_dollars)) print(" so the number the alert watches : cannot move") print("") print(" the alert is correctly wired to a quantity that this") print(" contract holds constant") print("") print("the commitment") print(" taken deliberately : yes, against on-demand pricing") print(" did it save money at the time : yes") print(" is it the mistake : no") print(" what it does to spend : fixes it for " + str(commitment_months) + " months") print(" what it does to the alert : leaves it correctly wired") print(" to a constant") print("") print("the migration") print(" moved the workload to : something that suits it better") print(" was that right : yes, on its own terms") print(" capacity needed after : lower") print(" capacity bought after : the same") print(" utilisation, before and after : " + str(utilisation_percent_before) + " and " + str(utilisation_percent_after)) print(" what appeared on the invoice : nothing") print("") print("utilisation") print(" is it measured : yes, the provider reports it") print(" is it on a dashboard : yes") print(" alerts configured on it : " + str(alerts_configured_on_utilisation)) print(" what it would have shown : " + str(utilisation_percentage_points_lost) + " points, in one month") print(" what nobody was watching for : a number going down") print(" dollars behind those points, so far : " + str(dollars_for_unused_capacity_since)) print("") nc_alerts_configured_on_utilisation = 2 nc_alerts_that_would_have_fired = 1 print("null control - a threshold on utilisation too") print(" spend thresholds : " + str(alert_thresholds_configured) + ", unchanged") print(" utilisation thresholds : " + str(nc_alerts_configured_on_utilisation)) print(" alerts that fire in the migration month : " + str(nc_alerts_that_would_have_fired)) print(" the alerting did not get better; it acquired a threshold") print(" on a quantity this contract lets move") print("") print("what a working budget alert guarantees") print(" spend crossing a threshold is noticed : exactly, tested") print(" with a real charge and acknowledged by a person") print(" waste is noticed : not addressed; the alert is") print(" denominated in spend, and this contract decouples") print(" spend from consumption") print("") print("an alert can only report on a quantity that varies; a") print("contract that fixes the monitored quantity does not break") print("the alert, it removes the alert's subject, and a silent") print("alert is what correct configuration looks like") print("") print("The alerting is done properly: " + str(alert_thresholds_configured) + " thresholds rather than one, tested by pushing a") print("real charge through, routed to " + str(named_people_who_acknowledge) + " named person who acknowledges. It fires on") print("spend, which a " + str(commitment_months) + "-month commitment holds at " + str(month_over_month_spend_variance_dollars) + " variance, so utilisation") print("falling from " + str(utilisation_percent_before) + " to " + str(utilisation_percent_after) + " percent produced " + str(alerts_that_fired) + " alerts while " + str(monthly_dollars_for_unused_capacity)) print("dollars a month bought nothing - " + str(dollars_for_unused_capacity_since) + " over " + str(months_since_the_migration) + " months.") ``` ## stdout (executed) ```text alert thresholds configured : 2 synthetic charge tests : 1 named people who acknowledge : 1 alerts that fired : 0 commitment months : 36 monthly commitment, dollars : 420000 month over month spend variance : 0 utilisation before, percent : 82 utilisation after, percent : 31 points lost : 51 unused after, percent : 69 monthly dollars for unused capacity : 289800 months since the migration : 7 dollars since : 2028600 alerts configured on utilisation : 0 the budget alerts thresholds : 2, so there is a warning before the wall rather than only a wall tested by : pushing a synthetic charge and watching it arrive, not by reading the console routed to : a named person, not a list acknowledged : yes verdict : CONFIGURED AND WORKING testing the alert with a real charge is the step that separates this from a checkbox, and it was done the monitored quantity what fires the alert : spend crossing a threshold what determines spend : the commitment, for 36 months what changes when capacity goes unused : nothing on the invoice month over month variance : 0 so the number the alert watches : cannot move the alert is correctly wired to a quantity that this contract holds constant the commitment taken deliberately : yes, against on-demand pricing did it save money at the time : yes is it the mistake : no what it does to spend : fixes it for 36 months what it does to the alert : leaves it correctly wired to a constant the migration moved the workload to : something that suits it better was that right : yes, on its own terms capacity needed after : lower capacity bought after : the same utilisation, before and after : 82 and 31 what appeared on the invoice : nothing utilisation is it measured : yes, the provider reports it is it on a dashboard : yes alerts configured on it : 0 what it would have shown : 51 points, in one month what nobody was watching for : a number going down dollars behind those points, so far : 2028600 null control - a threshold on utilisation too spend thresholds : 2, unchanged utilisation thresholds : 2 alerts that fire in the migration month : 1 the alerting did not get better; it acquired a threshold on a quantity this contract lets move what a working budget alert guarantees spend crossing a threshold is noticed : exactly, tested with a real charge and acknowledged by a person waste is noticed : not addressed; the alert is denominated in spend, and this contract decouples spend from consumption an alert can only report on a quantity that varies; a contract that fixes the monitored quantity does not break the alert, it removes the alert's subject, and a silent alert is what correct configuration looks like The alerting is done properly: 2 thresholds rather than one, tested by pushing a real charge through, routed to 1 named person who acknowledges. It fires on spend, which a 36-month commitment holds at 0 variance, so utilisation falling from 82 to 31 percent produced 0 alerts while 289800 dollars a month bought nothing - 2028600 over 7 months. ``` ## Round-trip `ok: true` — round-trip fixpoint reached (python1 == python2) ## Trace event types eml:run:start · eml:assign · eml:output · eml:run:done