CTO Systems Architecture & Infrastructure 2 — Questions and Answers
Question 1: A CTO is designing a multi-region deployment for a financial application that requires data residency compliance. Which architecture pattern best addresses both resilience and regulatory constraints?
- Active-active globally distributed with no data replication
- Active-passive with data pinned to compliant regions and cross-region failover for non-sensitive services (Correct answer)
- Single-region deployment with global CDN for all data
- Multi-cloud with unrestricted cross-border data sync
Correct answer: Active-passive with data pinned to compliant regions and cross-region failover for non-sensitive services
Active-passive with region-pinned data satisfies residency laws while still providing failover capability for the application tier.
Question 2: When evaluating a service mesh for a large microservices platform, which feature most directly improves observability without requiring code changes to individual services?
- Automatic certificate rotation
- Mutual TLS between services
- Distributed tracing via sidecar proxy injection (Correct answer)
- Circuit breaker configuration in application code
Correct answer: Distributed tracing via sidecar proxy injection
Sidecar proxies (e.g., Envoy in Istio) intercept traffic and emit traces automatically, providing observability without modifying service code.
Question 3: Your infrastructure team proposes replacing a monolithic database with a polyglot persistence strategy. What is the primary architectural risk the CTO must address first?
- Increased hardware costs
- Loss of ACID transactions that previously spanned multiple domains (Correct answer)
- Vendor lock-in on a single database engine
- Slower query performance on relational data
Correct answer: Loss of ACID transactions that previously spanned multiple domains
Polyglot persistence breaks cross-cutting transactions; the CTO must design compensating transactions or sagas to maintain data consistency.
Question 4: A startup CTO must choose between a shared-nothing and a shared-disk cluster architecture for an OLTP workload expected to scale to thousands of concurrent users. Which is the correct trade-off statement?
- Shared-disk scales write throughput better than shared-nothing
- Shared-nothing eliminates network overhead at the cost of requiring data partitioning logic (Correct answer)
- Shared-disk avoids single points of failure inherently
- Shared-nothing requires a central lock manager to coordinate writes
Correct answer: Shared-nothing eliminates network overhead at the cost of requiring data partitioning logic
Shared-nothing nodes own their data partitions, scaling linearly, but the application or middleware must handle sharding and cross-shard queries.
Question 5: Which infrastructure-as-code practice most reduces configuration drift in a fleet of 500 production servers?
- Running ad-hoc SSH commands logged to a change ticket
- Using immutable infrastructure with image-baked configuration and blue-green deployments (Correct answer)
- Applying Ansible playbooks manually before each release
- Storing server configs in a shared NFS mount
Correct answer: Using immutable infrastructure with image-baked configuration and blue-green deployments
Immutable infrastructure replaces servers rather than mutating them, eliminating drift by construction since no in-place changes are made post-deploy.
Question 6: A CTO reviews a proposal to use eventual consistency for a user-account balance service. What is the most significant reason to reject this proposal?
- Eventual consistency requires more storage capacity
- Balance reads could return stale data, enabling double-spend or overdraft scenarios (Correct answer)
- Eventual consistency cannot be implemented in cloud environments
- It prevents horizontal scaling of the database tier
Correct answer: Balance reads could return stale data, enabling double-spend or overdraft scenarios
Financial balances require strong consistency; eventual consistency windows create opportunities for incorrect reads that violate correctness guarantees for monetary values.
Question 7: When sizing a Kubernetes cluster for production workloads, which factor most often leads to node over-provisioning by engineering teams?
- Underestimating container image sizes
- Failing to account for Kubernetes system pod resource reservations per node (Correct answer)
- Using too many namespaces
- Setting PodDisruptionBudgets too aggressively
Correct answer: Failing to account for Kubernetes system pod resource reservations per node
System pods (kube-proxy, kubelet, CNI plugin, etc.) reserve CPU and memory on every node, reducing allocatable capacity below raw node specs.
CTO Systems Architecture & Infrastructure 2
A CTO is designing a multi-region deployment for a financial application that requires data residency compliance.
Which architecture pattern best addresses both resilience and regulatory constraints?