Zum Inhalt springen
Beta

The seven pillars

Zuletzt aktualisiert am

A workload is never optimal in every direction at once. Making it more reliable costs money. Making it more secure costs operational convenience. Making it faster usually costs both. The pillars name the qualities you are trading against each other, so that the trade is a decision rather than an accident.

Swipe sideways to see the whole diagram
How the seven pillars trade against each other Every pillar has a documented tradeoff against every other one: twenty-one pairs, each described from both sides. Select a pillar to see its six. Every pillar trades against every other one.Show what Operational Excellence trades againstOPSOperational ExcellenceShow what Security trades againstSECSecurityShow what Reliability trades againstRELReliabilityShow what Performance Efficiency trades againstPERFPerformance EfficiencyShow what Cost Optimization trades againstCOSTCost OptimizationShow what Sustainability trades againstSUSSustainabilityShow what Sovereignty & Compliance trades againstSOVSovereignty & ComplianceOpen Operational ExcellenceOPSOperational ExcellenceWhat pursuing it costs the other six:Operational Excellence against Security: Operations wants access; security wants it constrainedSECSecurityOperations wants access; security wantsit constrainedOperational Excellence against Reliability: Change is the leading cause of incidentsRELReliabilityChange is the leading cause of incidentsOperational Excellence against Performance Efficiency: Small and mostly mechanicalPERFPerformance EfficiencySmall and mostly mechanicalOperational Excellence against Cost Optimization: The most common conflict, and the one where "later" is most often the answerCOSTCost OptimizationThe most common conflict, and the onewhere "later" is most often the answerOperational Excellence against Sustainability: Minor, and it runs both waysSUSSustainabilityMinor, and it runs both waysOperational Excellence against Sovereignty & Compliance: Key lifecycle is an operating commitmentSOVSovereignty & ComplianceKey lifecycle is an operating commitmentOpen the full OPS tradeoffs pageAll OPS tradeoffsBack to the overviewClear selectionOpen SecuritySECSecurityWhat pursuing it costs the other six:Security against Operational Excellence: The dominant tension, and the source of most silently abandoned controlsOPSOperational ExcellenceThe dominant tension, and the source ofmost silently abandoned controlsSecurity against Reliability: Every control is a component that can failRELReliabilityEvery control is a component that canfailSecurity against Performance Efficiency: Encryption in transitPERFPerformance EfficiencyEncryption in transitSecurity against Cost Optimization: Security costs money in ways that are hard to attributeCOSTCost OptimizationSecurity costs money in ways that arehard to attributeSecurity against Sustainability: MinorSUSSustainabilityMinorSecurity against Sovereignty & Compliance: Mostly aligned, and the alignment hides a real divergenceSOVSovereignty & ComplianceMostly aligned, and the alignment hides areal divergenceOpen the full SEC tradeoffs pageAll SEC tradeoffsBack to the overviewClear selectionOpen ReliabilityRELReliabilityWhat pursuing it costs the other six:Reliability against Operational Excellence: Every reliability mechanism is something to operateOPSOperational ExcellenceEvery reliability mechanism is somethingto operateReliability against Security: Backups are copies of your dataSECSecurityBackups are copies of your dataReliability against Performance Efficiency: Synchronous cross-zone replicationPERFPerformance EfficiencySynchronous cross-zone replicationReliability against Cost Optimization: Redundancy against spendCOSTCost OptimizationRedundancy against spendReliability against Sustainability: Idle standby capacity consumes resources for no delivered workSUSSustainabilityIdle standby capacity consumes resourcesfor no delivered workReliability against Sovereignty & Compliance: Cross-region redundancy is a data placement decisionSOVSovereignty & ComplianceCross-region redundancy is a dataplacement decisionOpen the full REL tradeoffs pageAll REL tradeoffsBack to the overviewClear selectionOpen Performance EfficiencyPERFPerformance EfficiencyWhat pursuing it costs the other six:Performance Efficiency against Operational Excellence: The cost that gets underestimated mostOPSOperational ExcellenceThe cost that gets underestimated mostPerformance Efficiency against Security: Small and routinely overestimated in design discussionsSECSecuritySmall and routinely overestimated indesign discussionsPerformance Efficiency against Reliability: ScalingRELReliabilityScalingPerformance Efficiency against Cost Optimization: HeadroomCOSTCost OptimizationHeadroomPerformance Efficiency against Sustainability: The pillar with the most interesting relationshipSUSSustainabilityThe pillar with the most interestingrelationshipPerformance Efficiency against Sovereignty & Compliance: Placement constraintsSOVSovereignty & CompliancePlacement constraintsOpen the full PERF tradeoffs pageAll PERF tradeoffsBack to the overviewClear selectionOpen Cost OptimizationCOSTCost OptimizationWhat pursuing it costs the other six:Cost Optimization against Operational Excellence: Cost pressure against operationsOPSOperational ExcellenceCost pressure against operationsCost Optimization against Security: Security costs are easy to cutSECSecuritySecurity costs are easy to cutCost Optimization against Reliability: Redundancy against spendRELReliabilityRedundancy against spendCost Optimization against Performance Efficiency: HeadroomPERFPerformance EfficiencyHeadroomCost Optimization against Sustainability: Reserved and committed capacitySUSSustainabilityReserved and committed capacityCost Optimization against Sovereignty & Compliance: Sovereignty controls cost moneySOVSovereignty & ComplianceSovereignty controls cost moneyOpen the full COST tradeoffs pageAll COST tradeoffsBack to the overviewClear selectionOpen SustainabilitySUSSustainabilityWhat pursuing it costs the other six:Sustainability against Operational Excellence: Two directionsOPSOperational ExcellenceTwo directionsSustainability against Security: One structural conflict, and it is the subject of SUS 4.4SECSecurityOne structural conflict, and it is thesubject of SUS 4.4Sustainability against Reliability: Redundancy against consumptionRELReliabilityRedundancy against consumptionSustainability against Performance Efficiency: A genuinely two-sided relationshipPERFPerformance EfficiencyA genuinely two-sided relationshipSustainability against Cost Optimization: Committed capacityCOSTCost OptimizationCommitted capacitySustainability against Sovereignty & Compliance: Sovereignty wins these outright, and the conflicts are minorSOVSovereignty & ComplianceSovereignty wins these outright, and theconflicts are minorOpen the full SUS tradeoffs pageAll SUS tradeoffsBack to the overviewClear selectionOpen Sovereignty & ComplianceSOVSovereignty & ComplianceWhat pursuing it costs the other six:Sovereignty & Compliance against Operational Excellence: The largest cost, and the least anticipatedOPSOperational ExcellenceThe largest cost, and the leastanticipatedSovereignty & Compliance against Security: Aligned in mechanism, divergent in questionSECSecurityAligned in mechanism, divergent inquestionSovereignty & Compliance against Reliability: Key custody is a durability riskRELReliabilityKey custody is a durability riskSovereignty & Compliance against Performance Efficiency: Customer-managed keysPERFPerformance EfficiencyCustomer-managed keysSovereignty & Compliance against Cost Optimization: Confidential computingCOSTCost OptimizationConfidential computingSovereignty & Compliance against Sustainability: Small and mostly indirectSUSSustainabilitySmall and mostly indirectOpen the full SOV tradeoffs pageAll SOV tradeoffsBack to the overviewClear selectionTwenty-one pairs, all of them writtenEach pillar page carries a tradeoffs page stating what pursuing that pillarcosts the other six. Every one of the 21 pairs is described from both sides, 42sections in all.7pillars21pairs42sectionsSelect a pillar to see what it trades against.The selection stays until you pick another one.The framework names its own sharpest one: Security against Operational Excellence, "thedominant tension, and the source of most silently abandoned controls".

Each pillar page states what the pillar covers, what it deliberately does not, and the questions it asks, each with a summary. Beside it sit design principles (the reasoning), a page for each question (the detail), and a tradeoffs page (what pursuing this pillar costs the others).

Read the principles first if you are designing something, since that is where the reasoning lives and they are short. Go straight to the questions if you are reviewing something that exists.

Sovereignty is usually treated as a contractual matter, settled once in a procurement conversation and then assumed. On a sovereign cloud it is not. Where data resides, who can technically decrypt it, whether your logs leave the jurisdiction along with your data, whether you could migrate away if you had to: each of those is an architectural choice with real alternatives, and each is what an auditor and a customer questionnaire ask to see evidenced.

It does not decide whether your workload needs sovereignty, or how much. That is settled earlier, in the Advisory Framework, which classifies workloads into three sovereignty tiers. This pillar takes the tier as given and answers the next question: how you implement it, and how you would prove you did.

Questions are written REL 1, SEC 7, SOV 4. Where best practices are added beneath a question, they extend the identifier: REL 1.3. A compact form without the space, REL01, means the same thing, and is what appears in URLs and file names.

An identifier is permanent once published. A question that stops applying is retired rather than renumbered, and its number is not reused, so that a review recorded two years ago still means what it said.

Numbers do not indicate priority. They follow the order in which decisions are usually made, and new questions are appended rather than inserted, so even that ordering loosens over time.