SAP-C02 domain - 26% of the exam

Design Solutions for Organizational Complexity

Design Solutions for Organizational Complexity is 26% of the SAP-C02 exam. These are the objectives it covers, each with practice questions, with every answer explained.

The domain in numbers

72
Practice questions
8
Objectives
18
Medium
54
Hard

Where people go wrong

  • Using a permissions boundary where an SCP is needed

    The wrong answers offer a permissions boundary, or a deny written into one permission set, as the way to restrict accounts across the organisation. A permissions boundary caps a single principal, must be attached role by role, and a local administrator can create a role without it. A service control policy attached to the organisation root or an organisational unit sets the maximum permissions for every principal in the member accounts, so no account administrator can override it.

    See it in a sample question
  • Routing between VPCs when address ranges overlap

    The wrong answers route between VPCs where the requirement is to reach a service. Transit Gateway depends on unique IP ranges, and VPC peering cannot be created between VPCs that share a CIDR range, so two networks using 10.0.0.0/16 rule out both. AWS PrivateLink publishes a single service from a Network Load Balancer and each consumer reaches it through an interface endpoint, so overlapping ranges are irrelevant and nothing else in the VPC is exposed.

    See it in a sample question
  • Choosing a heavier DR tier than the targets need

    The wrong answers pick a recovery tier that costs more than the stated targets require. Active-active and warm standby keep a running fleet serving traffic, which is unnecessary when the recovery time objective is 15 minutes or several hours. A pilot light keeps only a cross-Region read replica and a minimal Auto Scaling group. Backup and restore stores snapshots and images, which fits a recovery point objective of a day.

    See it in a sample question
  • Counting same-location Direct Connect links as redundancy

    The wrong answers add capacity at the point that can still fail. A link aggregation group ends on the same device at one location, extra virtual interfaces share one physical connection, and two connections at a single location leave that location as one failure domain. Two dedicated connections at separate Direct Connect locations, on different customer routers with BGP across both, survive the loss of a device, a link or a location. Where the backup only carries reduced traffic briefly, a Site-to-Site VPN with BGP preferring Direct Connect costs less.

    See it in a sample question
  • Substituting AWS Config for CloudTrail or security findings

    The wrong answers reach for AWS Config where the requirement names a different record. Config records resource configuration state and rule compliance, not the management-event API calls a CloudTrail organisation trail captures, and a Config aggregator does not collect GuardDuty detections or Security Hub findings. An organisation trail created once in the management account covers every current and future member account.

    See it in a sample question

Objectives in this domain

What to study

The official documents this domain's practice questions cite most.

Sample questions from this domain

Free sampleDesign Solutions for Organizational Complexityhard

A platform team manages access with AWS IAM Identity Center across an organisation of 40 accounts. A single PowerUser permission set is assigned to many teams, but a recent review found it grants access to Amazon S3 buckets that several teams must never reach. Security wants every account that uses this permission set to be unable to call S3 regardless of what the permission set itself allows, while keeping a single change rather than editing dozens of roles. Which control MOST effectively enforces this organisation-wide boundary?

  • AEdit the PowerUser permission set in IAM Identity Center to add an explicit deny on all Amazon S3 actions, then let Identity Center automatically re-provision the updated role into each and every account where the permission set is currently assigned.
  • BAttach an IAM permission boundary that denies Amazon S3 to each role created by the permission set, replicating the boundary into all assigned accounts with a deployment pipeline so the limit travels with the role.
  • CApply a resource-based bucket policy on each Amazon S3 bucket that denies the principals provisioned by the permission set, listing their role ARNs so the buckets reject those callers across all of the accounts.
  • DCreate a service control policy that denies all Amazon S3 actions and attach it to the organisational units containing the affected accounts, so no principal in those accounts can call S3 even if an identity policy allows it. Correct
Choose a service control policy at the organisational unit to cap permissions org-wide, rather than per-role boundaries or per-resource policies. A service control policy defines the maximum available permissions for all IAM principals in the accounts it targets, including roles created by IAM Identity Center permission sets. A deny on S3 attached to the relevant organisational units therefore blocks the service everywhere in one change, which a permission-set edit, per-role boundary or per-bucket policy cannot guarantee across every current and future principal.

Why A is wrong: Adding a deny to the permission set does block S3 for that role and re-provisions everywhere, but it only constrains this one permission set, so any other role or a future permission set in those accounts could still reach S3.

Why B is wrong: A permission boundary can cap a principal's effective permissions, but it must be attached per role and maintained by a pipeline, which is heavier than one change and still misses any principal the boundary is not attached to.

Why C is wrong: Bucket policies only protect the specific buckets they sit on and require enumerating role ARNs, so new buckets and changed role names slip through and the control is not a true account-wide boundary.

Why D is correct: A service control policy sets the maximum permissions for every principal in the targeted accounts, so a deny on S3 at the organisational unit blocks the service org-wide in one place regardless of what any permission set or role grants.

Free sampleDesign Solutions for Organizational Complexityhard

A software vendor hosts a payments API behind a Network Load Balancer in its own VPC and account. Dozens of customer accounts, each with their own VPC and some using overlapping private CIDR ranges, must call only this single API over private networking, never reach anything else in the vendor VPC, and onboard without the vendor coordinating IP addressing with them. The vendor also wants no transitive path back into customer networks. Which approach BEST satisfies these constraints?

  • AAttach the vendor VPC and every customer VPC to a shared AWS Transit Gateway and use route tables so that customer attachments can route only to the subnet hosting the payments API in the vendor VPC.
  • BExpose the payments API as an AWS PrivateLink endpoint service fronted by the Network Load Balancer and let each customer create an interface endpoint to it, so they reach only that service over private connectivity. Correct
  • CCreate VPC peering connections from the vendor VPC to each customer VPC and add specific routes and security group rules so that only the payments API instances are reachable from each peered customer network.
  • DPublish the payments API as a Gateway VPC endpoint and have each customer create a gateway endpoint with a route table entry so traffic to the service stays on the AWS private network.
Use AWS PrivateLink to expose one private service across accounts with overlapping CIDRs and no transitive network path. PrivateLink publishes a single service behind a Network Load Balancer as an endpoint service, and consumers reach it through an interface endpoint that maps to a local ENI in their own VPC. Because connectivity is service-level and not IP routing, overlapping CIDR ranges do not matter, only the one service is reachable, and there is no transitive route back into either network, which peering and Transit Gateway cannot guarantee.

Why A is wrong: A Transit Gateway can scope routes per attachment, but it relies on IP routing, so overlapping customer CIDR ranges break it and it exposes a routed path into the vendor VPC rather than a single service.

Why B is correct: PrivateLink publishes the single API as an endpoint service, and each customer interface endpoint reaches only that service with no IP routing between VPCs, so overlapping CIDR ranges are irrelevant and nothing else is exposed.

Why C is wrong: Peering can restrict reachable hosts with routes and security groups, yet peering cannot be established between VPCs with overlapping CIDR ranges and still exposes IP-level reachability into the vendor VPC.

Why D is wrong: Gateway VPC endpoints exist only for Amazon S3 and DynamoDB, so they cannot expose a customer-built payments API and are not a mechanism for sharing a private service across accounts.

Free sampleDesign Solutions for Organizational Complexityhard

A media company runs a transactional web tier on an EC2 Auto Scaling group behind an Application Load Balancer and an Amazon RDS for PostgreSQL Multi-AZ database in eu-west-1. A new continuity standard mandates cross-Region recovery in a few minutes with a recovery point objective of seconds, while finance demands the smallest possible idle footprint in the recovery Region between disasters. The team is building a pilot light in eu-west-2 and must decide what to keep running there continuously. Which TWO components belong in the pilot light so it meets the targets at the leanest standing cost? Select TWO.

  • AA full-capacity Auto Scaling group in eu-west-2 already running production-sized instances and serving a share of live customer traffic alongside the primary Region at all times.
  • BNightly automated RDS snapshots copied to eu-west-2 from which a fresh database cluster is restored only after eu-west-1 is declared lost during the failover event.
  • CA cross-Region RDS for PostgreSQL read replica in eu-west-2 kept continuously in sync so the data layer can be promoted to a writer within minutes during a Regional failover. Correct
  • DPre-created core infrastructure in eu-west-2, the VPC, subnets, security groups and a launch template, with the Auto Scaling group at zero or minimal capacity ready to scale out on failover. Correct
  • EA reduced-capacity Auto Scaling group in eu-west-2 continuously serving a fraction of production traffic so warm instances are always available to absorb the failover load.
Identify the always-on data replica and pre-staged scaled-down compute as the two pilot light components meeting a minutes RTO and seconds RPO cheaply. A pilot light keeps the data layer replicating live for a low RPO and pre-provisions core infrastructure with compute scaled to near zero, so the recovery Region costs little when idle yet scales out within minutes when a failover is triggered.

Why A is wrong: Running full production compute serving live traffic is multi-site active-active, not a pilot light, and its standing cost far exceeds the lean idle footprint that finance demands here.

Why B is wrong: Nightly snapshot copies are a backup-and-restore tactic that loses up to a day of data and takes a long restore, so they cannot deliver the seconds RPO or the minutes RTO required.

Why C is correct: Continuous read-replica replication keeps the data layer current for a seconds-level RPO and promoting it on failover meets the minutes RTO, which is the core always-on element of a pilot light.

Why D is correct: Pre-staging the network and launch template with compute scaled to near zero lets the fleet expand quickly on failover while keeping idle cost minimal, which is exactly the pilot light compute posture.

Why E is wrong: An always-on reduced fleet serving live traffic is warm standby, which costs more than a pilot light needs and exceeds the leanest idle footprint the scenario requires.

Free sampleDesign Solutions for Organizational Complexityhard

A financial services firm connects its primary data centre to AWS for a settlement workload that needs predictable low latency and a sustained 5 Gbps of throughput that must not be affected by other tenants on the link. The network team also requires the connection to survive the loss of a single device or fibre path inside the data centre without dropping the workload. Which connectivity design MOST resiliently meets the dedicated-bandwidth and single-fault-tolerance requirements?

  • AProvision a single 10 Gbps AWS Direct Connect dedicated connection at one location and configure a Site-to-Site VPN over the public internet as the standby path to take over if the dedicated link fails.
  • BProvision two AWS Direct Connect dedicated connections from separate customer routers, each at a different Direct Connect location, and run BGP across both so traffic reroutes to the survivor on a fault. Correct
  • COrder two AWS Direct Connect dedicated connections terminating on separate devices at two Direct Connect locations and join them in a link aggregation group so traffic balances across both physical links.
  • DProvision one AWS Direct Connect dedicated connection and split it into two virtual interfaces on separate VLANs so each settlement traffic class has its own isolated path across the shared physical fibre run.
Achieve maximum hybrid resilience with two Direct Connect connections at separate locations on separate devices running BGP, not a LAG or virtual interfaces. AWS guidance for the highest resilience is two Direct Connect dedicated connections terminating on different customer devices at separate Direct Connect locations, with BGP providing automatic failover between them. A LAG cannot span locations, multiple virtual interfaces share one physical link, and an internet VPN cannot guarantee dedicated low-latency throughput, so only the dual-location dedicated design survives a single device or fibre fault.

Why A is wrong: A dedicated connection gives the bandwidth, but a single connection plus an internet VPN backup leaves a device or fibre fault able to drop the link, and the VPN cannot sustain a predictable 5 Gbps of low-latency throughput on failover.

Why B is correct: Two dedicated connections from different customer devices at separate Direct Connect locations give isolated dedicated bandwidth and, with BGP across both, automatically reroute the workload if any single device or fibre path fails, meeting the maximum-resilience guidance.

Why C is wrong: A link aggregation group is tempting for resilience, but a LAG requires all member connections to terminate on the same AWS device at one location, so it cannot span two locations and does not deliver true location-level fault tolerance.

Why D is wrong: Multiple virtual interfaces partition traffic logically but all ride the same physical connection, so a single device or fibre fault still takes the whole link down, providing no fault tolerance at all.

Free sampleDesign Solutions for Organizational Complexityhard

A financial services group runs about 40 accounts in a single AWS Organization with all features enabled. The security team must guarantee that every existing and future member account records management-event API activity to one immutable, central location that account administrators cannot disable or reconfigure, and they want the smallest amount of ongoing setup work per new account. Which approach MOST effectively delivers this tamper-resistant central audit trail?

  • AAsk each account owner to create an account-level CloudTrail trail delivering to a shared S3 bucket, and rely on a weekly Config rule to flag any account whose trail has been stopped or deleted by an administrator.
  • BEnable Amazon GuardDuty across the organisation from a delegated administrator account so that all API-derived threat findings are aggregated centrally, then export the GuardDuty findings to a locked S3 bucket for the auditors to review.
  • CTurn on AWS Config across all accounts with an aggregator in a central account, recording configuration item changes to a shared bucket, and treat the Config history as the authoritative organisation-wide API audit log for the auditors.
  • DCreate an organisation trail from the management account that applies to all member accounts, delivering to a central S3 bucket with Object Lock, and use an SCP to deny cloudtrail:StopLogging and trail deletion in member accounts. Correct
Combine a CloudTrail organisation trail, an Object Lock S3 bucket and an SCP to enforce an immutable, undisableable central audit trail across all accounts. A CloudTrail organisation trail is defined once in the management or delegated administrator account and automatically applies to every member account, including ones created later, so no per-account setup is needed. Delivering to an S3 bucket with Object Lock makes the stored logs immutable, and an SCP that denies cloudtrail:StopLogging and trail mutation prevents member administrators from disabling auditing. GuardDuty and Config aggregate other telemetry but do not produce the complete immutable API record.

Why A is wrong: Per-account trails plus a detective Config rule can surface drift, but they let a local administrator pause or delete logging between evaluations, so the trail is neither guaranteed-on nor truly central and adds work per account.

Why B is wrong: GuardDuty consumes CloudTrail and other telemetry to produce threat findings centrally, but it is a detection service and does not itself create the complete, immutable record of every management API call that the audit requirement demands.

Why C is wrong: Config records resource configuration state over time and aggregates it centrally, which looks audit-like, but it captures configuration changes rather than the full set of management-event API calls, so it cannot replace a central CloudTrail organisation trail.

Why D is correct: An organisation trail is created once and auto-enrols every current and future account, an Object Lock bucket makes the logs immutable, and a service control policy removes member administrators' ability to stop or alter logging, giving a tamper-resistant central trail with minimal per-account effort.

Other domains in this exam

See also the SAP-C02 cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by Amazon Web Services. Original, blueprint-aligned practice material only.