Cloud Computing Senior
What is the difference between Data Residency and Data Sovereignty, and why does a developer need to care about which Region their data is stored in?
Select the correct answer
Residency is the physical location where data is stored, while sovereignty means the data is subject to the laws of the country it resides in, so Region choice affects legal compliance
Residency is the set of laws governing the data, while sovereignty is simply the physical location of storage, so Region choice mainly affects network latency and replication throughput speeds
Residency and sovereignty both refer to encryption standards applied at rest, so Region choice only matters for selecting which key management service and cipher suite your application will use
Residency is a backup retention policy while sovereignty is an access control model, so Region choice determines how long snapshots persist and which IAM roles can read the data
What is 'Zero Trust' architecture and how does it differ from traditional perimeter-based security?
Select the correct answer
Zero Trust removes all authentication to speed up internal calls, whereas perimeter security relies purely on continuous identity verification for every user and device that connects
Zero Trust and perimeter security are identical models, both trusting internal traffic while blocking external requests using a hardened firewall placed at the edge of the corporate network
Zero Trust assumes no request is inherently trusted and verifies every one, whereas perimeter security trusts anything already inside the network boundary once it passes the firewall
Zero Trust trusts all internal traffic to reduce overhead, whereas perimeter security individually authenticates every single request regardless of whether it originates inside or outside the network
What are 'The Nines' in an SLA (e.g., 99.9% vs 99.99%), and how does increasing the number of nines impact the cost and complexity of your architecture?
Select the correct answer
They express the guaranteed uptime percentage; adding a nine increases allowed downtime, so it lets you cut redundancy and reduces both the cost and complexity of the design
They express the average request latency in milliseconds; adding a nine tightens the latency budget, so it requires faster hardware but has little effect on redundancy or cost
They express the percentage of data durability guaranteed; adding a nine improves replication factors, so it mainly affects storage spend without changing availability or overall complexity
They express the guaranteed uptime percentage; adding a nine shrinks allowed downtime dramatically, so it demands more redundancy and raises cost and complexity significantly
What is 'Immutable Infrastructure,' and how does it differ from traditional server maintenance ('Pets vs. Cattle')?
Select the correct answer
Servers are continuously patched and updated in place after deployment, treating each one as a unique disposable unit that is rebuilt from scratch on every configuration change
Servers are manually configured once and then frozen from all monitoring, treating them as irreplaceable pets that must be individually backed up before any code deployment
Servers are never modified after deployment; changes mean replacing instances entirely, treating them as disposable cattle rather than carefully patched, hand-nursed pets
Servers share a single mutable image edited live across the fleet, treating the whole cluster as one pet that is upgraded gradually to avoid replacing running instances
What is key management in the cloud, and what are the trade-offs between provider-managed keys and customer-managed keys?
Select the correct answer
Provider-managed keys cannot be rotated, while customer-managed keys are automatically rotated by the cloud vendor
Provider-managed keys give full customer control, while customer-managed keys are handled entirely by the vendor
Provider-managed keys are simpler to operate; customer-managed keys give more control but add operational burden
Provider-managed keys are always less secure, while customer-managed keys remove all compliance responsibility
What does it mean to 'Design for Failure' in the cloud? Give an example of a common cloud failure and how you would mitigate it conceptually.
Select the correct answer
Purchase premium hardware and support plans so failures never occur in production.
Assume components will fail and build redundancy so the system recovers automatically.
Prevent outages entirely by running the whole application in a single strong region.
Eliminate all failures by testing code thoroughly before deploying it to production.
What are common disaster recovery strategies (backup-and-restore, pilot light, warm standby, active-active), and how do they trade cost against RTO/RPO?
Select the correct answer
Cost is unrelated to RTO/RPO because each strategy relies on the same standby capacity and differs only in the region chosen
Cheaper strategies like active-active give the lowest RTO/RPO, while pricier backup-and-restore trades speed for stronger data durability
Higher spend on always-running standby infrastructure buys higher RTO/RPO, moving from active-active down to plain backup-and-restore
Higher spend on always-running standby infrastructure buys lower RTO/RPO, moving from backup-and-restore up to active-active
What are 'Egress Costs,' and why are they often referred to as a 'hidden' cloud cost? How can a developer architect a system to minimize them?
Select the correct answer
Charges for data entering the cloud; route all inbound uploads through a single region and disable caching to reduce transfers
Charges for storing archived data long-term; move objects to colder storage tiers frequently to reduce the recurring transfer fees
Charges for API request counts on services; batch calls together and add more compute nodes to reduce the per-request transfer fees
Charges for data leaving the cloud; keep traffic within the same region and use caching or CDNs to reduce transfers
Explain the concept of 'Total Cost of Ownership' (TCO) in the cloud. Why might a cloud bill be higher than the cost of the physical servers it replaced?
Select the correct answer
TCO counts only upfront hardware purchases; bills grow because providers must recover the cost of the datacenter buildings they own
TCO counts only the monthly compute charge; bills grow because cloud CPUs are always inherently slower than owned physical hardware
TCO counts staffing salaries alone; bills grow chiefly because cloud contracts require paying for unused capacity you never provisioned
TCO counts all direct and indirect costs; bills grow from data transfer, managed services, and idle over-provisioned resources
What is FinOps, and what do 'showback' and 'chargeback' mean in cloud cost management?
Select the correct answer
FinOps is a billing negotiation process with vendors; showback and chargeback both mean forecasting future costs based on historical consumption trends.
FinOps is a discipline for cloud cost accountability; showback reports each team's costs for visibility, while chargeback actually bills teams for their usage.
FinOps is a security compliance framework; showback bills each team directly, while chargeback only reports usage to teams for internal visibility purposes.
FinOps is an automated scaling technique; showback allocates unused credits to teams, while chargeback reserves capacity in advance to reduce total spend.
What is a 'Cold Start' in serverless computing, and what architectural patterns can mitigate its impact?
Select the correct answer
Latency from encrypting the first request; disabling TLS on internal endpoints avoids it completely.
Latency from a full server reboot on deploy; adding more memory to each function removes it entirely.
Latency caused by database connection pooling; caching every query result eliminates it across regions.
Latency to initialize a new function instance; provisioned concurrency or keep-warm calls reduce it.
Explain the '6 Rs' of cloud migration. When would you recommend a 'Lift and Shift' (Rehost) over a full Refactor?
Select the correct answer
Rehost when you want to fully re-architect the app to gain elasticity and cost savings
Rehost only when the application must be rebuilt to use managed cloud-native services first
Rehost when the legacy application should be retired and replaced with a SaaS product
Rehost when speed matters and the app can move as-is; refactor later once value is proven
What is 'Vendor Lock-in', and what architectural patterns can help mitigate it?
Select the correct answer
Dependence on a single billing contract; reserved instances and spot pricing help to reduce it
Dependence on one region's data centers; multi-zone replication and failover fully eliminate it
Dependence on legacy hardware vendors; virtualization and hypervisor tuning largely eliminate it
Dependence on one provider's proprietary services; abstraction layers and open standards reduce it
Explain the difference between Hybrid Cloud and Multi-Cloud. Why might a company choose a multi-cloud strategy despite the increased complexity?
Select the correct answer
Hybrid mixes private and public; multi-cloud uses several public providers to avoid lock-in
Hybrid uses several public providers; multi-cloud mixes private and public to boost security
Hybrid means SaaS plus IaaS; multi-cloud means running only containers across data centers
Hybrid and multi-cloud both mean one provider spanning many regions for lower latency
What is the difference between a 'Cloud-Native' and a 'Cloud-Agnostic' architecture? What are the risks of vendor lock-in?
Select the correct answer
Cloud-native means virtual machines only; cloud-agnostic means serverless functions on any provider
Cloud-native means on-premises only; cloud-agnostic means using exclusively public cloud offerings
Cloud-native stays portable across providers; cloud-agnostic embraces one provider's managed services
Cloud-native embraces one provider's managed services; cloud-agnostic stays portable across providers