CCNA Cloud Architecture and Design Questions

Serverless computing (FaaS) bills only for actual invocation and execution duration, measured in milliseconds, rather than for idle provisioned capacity. This consumption-based model directly satisfies the characteristic sought: no charges accrue while code sits dormant, unlike always-on virtual machines or containers.

RDS Multi-AZ maintains a synchronous standby in another Availability Zone and automatically promotes it during a failure, updating the DNS endpoint so the application reconnects without manual intervention. This satisfies the database requirement for automatic failover with minimal administration.

Multi-cloud means deliberately using two or more distinct public cloud providers, here AWS and Azure, for the same workload. This satisfies the redundancy, lock-in avoidance and disaster recovery constraints, unlike hybrid cloud, which pairs public cloud with private infrastructure.

File storage exposes SMB or NFS shares that many VMs can mount concurrently with read/write access, satisfying the simultaneous multi-mount constraint. Block storage attaches to a single instance, and object storage uses HTTP APIs rather than mountable file protocols.

A Kubernetes Service provides a stable virtual IP and DNS name that abstracts a set of pods. It enables dynamic discovery and load balancing across pods, even as they are created or destroyed. This matches the requirement for a stable network endpoint without hardcoded IPs.

Cross-zone load balancing distributes traffic evenly across healthy targets in every availability zone, and health checks withdraw targets in a failed zone automatically. This satisfies the stem's zone-failure tolerance constraint by removing the failed zone from rotation without manual intervention.

S3 Object Lock in compliance mode enforces a write-once-read-many retention period that even the root user cannot shorten or bypass, and objects cannot be overwritten or deleted until the date passes. Setting a 2555-day retention on the bucket satisfies the exact seven-year, tamper-proof requirement for every stored video master.

IaaS leaves the customer managing the guest operating system, middleware and applications, while the provider handles physical hardware, networking and storage virtualisation. This split gives more control than PaaS or SaaS, which abstract the OS layer away.

Multi-cloud means deliberately using two or more distinct public cloud providers for different workloads. AWS for compute plus Azure for storage spans separate providers, satisfying the best-of-breed requirement rather than a hybrid or single-provider model.

Reserved instances apply a one- or three-year commitment to steady-state capacity, cutting hourly rates substantially versus on-demand. This satisfies the cost optimisation requirement by discounting the predictable baseline load while on-demand or spot covers peaks.

Spot instances price capacity at steep discounts because workloads are evictable, matching the job's interruptibility. Reserved or on-demand pricing would charge for guaranteed availability the batch job does not require, so spot satisfies the nightly two-hour, interruption-tolerant constraint most cheaply.

Active-active across two Availability Zones runs identical workloads in both, so each AZ handles traffic independently. When one AZ fails, the surviving AZ absorbs the load, satisfying the requirement to keep serving traffic through a full AZ outage.

Azure ExpressRoute provides a dedicated, private circuit through a connectivity provider, bypassing the public internet entirely. This satisfies the stem's requirement for private, high-bandwidth connectivity between the on-premises data centre and Azure, unlike VPN gateways which traverse the public internet and offer variable bandwidth.

Direct Connect provides a dedicated private connection with consistent performance.

A shared database externalises all session and user state, so any instance can serve any request without relying on local memory. This satisfies the horizontal-scaling constraint, since instances remain interchangeable and can be added or removed freely.

Hybrid cloud joins on-premises infrastructure to a public cloud through VPN tunnels or dedicated private links such as AWS Direct Connect, keeping workloads split across both environments. That connectivity mechanism is exactly what the stem describes.

Cloud CDN caches static assets at Google's global edge points of presence, cutting latency for images and CSS by serving users from nearby locations rather than the origin. It directly satisfies the stem's low-latency static-content constraint. SSL termination, however, is handled separately by the external Application Load Balancer, not Cloud CDN itself.

Block storage presents a raw volume directly to the VM, giving the database low-latency, consistent read/write performance on a single host. File and object storage add network protocol overhead and are unsuitable for database workloads needing predictable IOPS, so block satisfies the single-VM, consistent-performance constraint.

Amazon S3 provides high durability and availability for unstructured data. Cross-Region Replication automatically replicates objects to another region, enabling low-latency access from multiple regions. S3 Lifecycle policies can transition infrequently accessed data to S3 Glacier Deep Archive for long-term retention at low cost. This combination meets all requirements.

Configuring auto-scaling lets the PaaS platform add and remove compute instances automatically as demand rises, directly meeting the no-downtime requirement under increased load. Because the provider manages the underlying infrastructure, scaling occurs without manual intervention or service interruption, unlike vertical resizing, which typically needs a restart.

Moving session state to a shared database or cache externalises it from individual instances, so any server behind the load balancer can serve any request. This directly satisfies the horizontal scaling requirement, since locally stored sessions would otherwise pin users to one server and break scaling.

A managed Kubernetes service provides a control plane that schedules containers, restarts failed pods through the ReplicaSet controller, and abstracts the worker nodes. Horizontal Pod Autoscalers adjust replica counts per service based on metrics, so each microservice scales independently. Because the provider manages the control plane, the operations team's limited cluster expertise is less of an obstacle.

Read replicas offload read traffic from the primary and can be added or removed while the database stays online. A reader endpoint or proxy presents a stable address, so the application keeps the same connection string while read capacity scales horizontally. This directly addresses unpredictable read spikes without downtime or application changes.

An organization trail automatically applies to all accounts in the organization, including accounts added later, and delivers events to a central S3 bucket. The SCP then removes the ability of member account principals to stop or delete the trail, satisfying the tamper-resistance requirement without per-account configuration.

High availability across two Availability Zones requires each subnet tier duplicated per AZ, giving two public and two private subnets. Public subnets host the internet-facing web servers via an internet gateway, while private subnets hold the databases without inbound internet routing, satisfying both the accessibility and isolation constraints.

Multi-cloud means consuming services from two or more public cloud providers simultaneously, distributing workloads across them. This avoids dependence on a single vendor and lets each workload use the strongest provider service, directly matching the stated vendor lock-in and best-of-breed requirements.

Spot instances use spare capacity priced far below on-demand rates, and their interruptible nature suits a nightly batch job that can tolerate eviction and resume. This directly satisfies the stem's cost-minimisation and interruption-tolerance constraints, unlike reserved or on-demand pricing.

Amazon EBS Provisioned IOPS SSD (io2 Block Express) is designed for critical, I/O-intensive database workloads that require the highest levels of IOPS and consistently low latency. It supports up to 256,000 IOPS and 4,000 MB/s per volume with sub-millisecond latency, and offers 99.999% durability. It is the best choice for a single-instance database requiring maximum performance.

A CDN caches static assets at edge locations, cutting latency for images and CSS. Dynamic API responses cannot be cached, so a global load balancer with latency-based routing directs each user to the nearest healthy regional endpoint, satisfying the worldwide latency constraint.

A global external Application Load Balancer uses a single anycast IP address and routes user traffic to the closest healthy backend based on Google's global network. It operates at layer 7, supports HTTP(S) workloads, and is designed for global low-latency distribution. This matches the requirement for a single IP and proximity-based routing across regions.

Amazon EFS is a fully managed, scalable, and highly available file storage service for Linux-based workloads. It supports the NFS protocol, can be mounted on multiple EC2 instances concurrently, and supports POSIX permissions. This meets all the requirements for a shared file system.

Azure Files provides fully managed SMB and NFS file shares, so multiple Linux VMs can mount the same share concurrently using standard protocols. This directly satisfies the stem's requirement for shared storage across Linux VMs, unlike blob storage or single-VM managed disks, which lack native multi-attach file sharing.

Archive storage offers the lowest per-gigabyte cost of the tiers, designed for data retained for years with infrequent access. The stem states retrieval time is not critical, so the slower, pricier-retrieval trade-off is acceptable for seven-year compliance retention.

Azure Files exposes SMB and NFS shares, so on-premises servers and cloud VMs mount the same file system concurrently. This satisfies the hybrid shared-file-system constraint without replication or application changes, unlike object storage such as Blob.

Continuous transaction log shipping replicates every committed transaction to a standby database, achieving an RPO of near zero or within minutes. If the primary fails, the standby can be promoted, and recovery time is typically well under an hour. This meets both the 15-minute RPO and the 1-hour RTO, providing a robust disaster recovery solution.

Asynchronous replication copies changes after acknowledgement, so a site can lag by up to an hour — matching the one-hour RPO — while costing far less than synchronous replication, which would need low-latency links to meet a near-zero RPO the business does not require.

Stateless application design ensures no session or local data persists on individual instances, so any instance can handle any request. This removes the need for sticky sessions or instance-specific state, directly enabling horizontal scaling and safe termination during scale-in events.

File storage provides a shared, POSIX-compliant file system mountable concurrently by multiple Linux instances within the same VPC, matching the requirement. Block storage attaches to a single instance, and object storage lacks native shared-mount semantics.

Preemptible VMs are cost-effective and can be interrupted, ideal for batch jobs.

Hybrid cloud keeps sensitive customer data on-premises under the firm's control while connecting to public cloud resources for peak risk-analysis bursts. This directly satisfies both constraints: data residency on-premises and elastic compute scaling during peak demand.

A NAT Gateway allows instances in a private subnet to initiate outbound traffic to the internet while preventing inbound traffic from reaching them. Placing the NAT Gateway in a public subnet gives it internet access via the Internet Gateway. Updating the private subnet's route table to point to the NAT Gateway enables the application servers to download updates without being directly accessible.

Active-active across availability zones with a load balancer satisfies the automatic failover constraint: the load balancer health-checks each zone's endpoints and reroutes traffic to surviving zones when one fails, with no manual intervention. Running identical instances in every zone also preserves capacity, unlike active-passive, which requires failover orchestration.

Auto-failover groups provide a read-write listener endpoint that remains constant during failover, enabling automatic failover of one or more databases to a secondary region. They support automatic replication and can meet low RPO and RTO requirements by handling the failover process seamlessly.

Bursting to the cloud directly satisfies the stem's requirement to leverage cloud resources for extra capacity while sensitive data stays on-premises. In a hybrid model, workloads scale out to public cloud during demand spikes, then return on-premises, preserving regulatory compliance without over-provisioning internal hardware.

In-memory session state binds each user to one server, preventing horizontal scaling. Moving sessions to a shared cache or database makes application instances stateless, so any instance can serve any request and additional replicas absorb traffic spikes.

Separating web, application and database tiers into public and private subnets, then restricting traffic with security group rules referencing each tier, enforces the required access path. Only web servers are internet-facing, satisfying the constraint that application and database servers remain unreachable directly.

AWS Lambda supports SQS as an event source. When configured, Lambda automatically polls the queue and invokes the function with batches of messages. It scales based on the number of messages, and failed invocations can be retried or sent to a dead-letter queue. This meets the requirements for automatic scaling and graceful failure handling.

An Auto Scaling group spanning multiple Availability Zones ensures that if one AZ fails, instances in the remaining AZ continue to serve traffic. The Auto Scaling group can also replace failed instances automatically. This provides compute-layer high availability and is a fundamental design practice for surviving AZ failures.

S3 Lifecycle policies automatically transition infrequently accessed datasets to Glacier, whose storage pricing is far lower than S3 Standard. This satisfies the cost-reduction goal for large, ageing analytics data without altering the EC2 compute tier.

Active-active deployment runs workloads simultaneously in both on-premises and cloud, so traffic fails over instantly when one site degrades. This directly satisfies the stem's high-availability requirement for a critical application, unlike active-passive designs that incur downtime during failover. Distributing load across both environments also removes the single point of failure inherent in a single-site deployment.

A cloud-agnostic orchestration tool such as Terraform or Kubernetes abstracts provider-specific APIs, letting one toolset manage AWS and Azure workloads. This directly satisfies the requirement to simplify management across both clouds while preserving the redundancy architecture.

A hybrid cloud specifically joins on-premises infrastructure to a public cloud, typically via VPN or dedicated private link. This satisfies the stem's requirement for that on-premises-to-public-cloud connectivity, distinguishing it from purely public, private or community models.

ARM templates describe the desired infrastructure declaratively, so the same template can be redeployed to a secondary region to recreate virtual networks, load balancers, and VMs consistently. This provides the automated, repeatable regional deployment the team needs and supports parameterization for region-specific values, making it the appropriate choice for this recovery scenario.

A warm standby runs a reduced but functional copy of the application in another region, so failover does not require building infrastructure from scratch. It can be scaled to full capacity when the primary region fails, meeting the automated, low-downtime requirement while respecting the budget that rules out a fully active-active deployment. This directly addresses surviving a regional outage.

Latency routing directs each user to the region responding fastest, measured by actual network round-trip time. This satisfies the minimise-latency constraint for global users, unlike failover or weighted policies which prioritise availability or distribution rather than per-user responsiveness.

Hybrid cloud combines on-premises infrastructure with public cloud services, letting sensitive applications remain on local hardware while other workloads migrate. This directly satisfies the stem's dual requirement: cloud migration plus on-premises retention, which neither public nor private cloud alone can deliver.

Network tags are used in conjunction with VPC firewall rules to identify source or target instances. By tagging instances in each tier, you can create firewall rules that allow traffic only from specific tags, enabling granular control over which tiers can communicate, thus enforcing the desired segmentation.

Direct Connect (AWS) and ExpressRoute (Azure) provision private, dedicated circuits from an on-premises data centre to the provider's edge, bypassing the public internet entirely. This satisfies the requirement for a consistent, dedicated connection rather than an IPsec VPN tunnel over internet links.