10 free sample questions from a bank of 272, with the correct answers and explanations. No signup required — start practising right now.
1A company is planning to create a service that requires encryption in transit. The traffic must not be decrypted between the client and the backend of the service. The company will implement the service by using the gRPC protocol over TCP port 443. The service will scale up to thousands of simultaneous connections. The backend of the service will be hosted on an Amazon Elastic Kubernetes Service (Amazon EKS) duster with the Kubernetes Cluster Autoscaler and the Horizontal Pod Autoscaler configured. The company needs to use mutual TLS for two-way authentication between the client and the backend.
Which solution will meet these requirements?
Install the AWS Load Balancer Controller for Kubernetes. Using that controller, configure a Network Load Balancer with a TCP listener on port 443 to forward traffic to the IP addresses of the backend service Pods.
Install the AWS Load Balancer Controller for Kubernetes. Using that controller, configure an Application Load Balancer with an HTTPS listener on port 443 to forward traffic to the IP addresses of the backend service Pods.
Create a target group. Add the EKS managed node group's Auto Scaling group as a target Create an Application Load Balancer with an HTTPS listener on port 443 to forward traffic to the target group.
Create a target group. Add the EKS managed node group’s Auto Scaling group as a target. Create a Network Load Balancer with a TLS listener on port 443 to forward traffic to the target group.
Answer: A
The short version
A — NLB with TCP passthrough preserves end-to-end mTLS for gRPC. It needs intact encryption from client to pod plus mTLS. Only a Layer 4 TCP forwarder keeps TLS untouched while tracking pod IPs as they churn.
Key concepts in this question
End-to-end TLS: load balancer forwards bytes without decrypting, so client and backend complete handshake directly.
gRPC over TCP 443: HTTP2 based RPC that needs transparent transport, not HTTP termination, to preserve streaming and trailers.
Mutual TLS: both sides present certificates, requiring the balancer to avoid breaking the chain with its own certificate.
EKS autoscaling: Cluster Autoscaler plus HPA constantly adds and removes pods, so targets must update by pod IP.
AWS Load Balancer Controller: Kubernetes controller that provisions NLB or ALB and syncs endpoints automatically.
Why A is correct
Option A uses NLB with a TCP listener on 443 pointing at pod IPs, so packets pass through without TLS termination and mTLS succeeds between client and EKS backend. TCP mode supports any TCP payload including gRPC, scales to thousands of concurrent flows, and preserves source transparency needed for mutual authentication. Because the controller registers pod IPs directly, scaling events update targets without manual target group work, meeting availability and automation needs in one design.
Why the others are wrong
B. uses ALB with HTTPS, which terminates TLS and re-encrypts, breaking the no-decrypt and mTLS requirement.
C. uses ALB via node Auto Scaling group targets, adding TLS break plus coarse instance routing instead of pod IP routing.
D. uses NLB with TLS listener, which still terminates TLS at the balancer and defeats end-to-end mutual authentication.
ANS-C01 exam tip
When the stem says must not be decrypted or needs mTLS end to end, choose TCP passthrough on NLB and reject any ALB HTTPS or TLS listener.
2A company is deploying a new application in the AWS Cloud. The company wants a highly available web server that will sit behind an Elastic Load Balancer. The load balancer will route requests to multiple target groups based on the URL in the request. All traffic must use HTTPS. TLS processing must be offloaded to the load balancer. The web server must know the user’s IP address so that the company can keep accurate logs for security purposes.
Which solution will meet these requirements?
Deploy an Application Load Balancer with an HTTPS listener. Use path-based routing rules to forward the traffic to the correct target group. Include the X-Forwarded-For request header with traffic to the targets.
Deploy an Application Load Balancer with an HTTPS listener for each domain. Use host-based routing rules to forward the traffic to the correct target group for each domain. Include the X-Forwarded-For request header with traffic to the targets.
Deploy a Network Load Balancer with a TLS listener. Use path-based routing rules to forward the traffic to the correct target group. Configure client IP address preservation for traffic to the targets.
Deploy a Network Load Balancer with a TLS listener for each domain. Use host-based routing rules to forward the traffic to the correct target group for each domain. Configure client IP address preservation for traffic to the targets.
Answer: A
The short version
A — ALB with HTTPS plus path routing and X-Forwarded-For meets all four needs. Layer 7 routing by URL path, centralized TLS offload and client IP preservation in logs point to one service. Only an ALB combines path conditions, HTTPS termination and header insertion natively.
Key concepts in this question
Path-based routing: ALB listener rules inspect URL path and send each prefix to a different target group from one balancer.
TLS offload: certificate lives on the balancer, backends receive plain HTTP, reducing compute on web servers.
X-Forwarded-For: header injected by ALB carrying original client IP when the TCP source becomes the balancer address.
Host versus path: host rules match domain names, path rules match URL segments, the stem explicitly asks for URL based split.
NLB limits: Network Load Balancer works at Layer 4 and cannot inspect HTTP path, only IP, port or SNI.
Why A is correct
Option A deploys ALB with an HTTPS listener, offloads TLS at the edge, uses path conditions to fan out to multiple target groups, and adds X-Forwarded-For so the web tier logs true client addresses. That single design satisfies highly available web fronting, URL driven routing to several groups, mandatory HTTPS, offload, and audit logging. One balancer with path rules is simpler and cheaper than per-domain balancers and matches AWS guidance for HTTP aware fan-out.
Why the others are wrong
B. uses per-domain host routing while the requirement is path based, adding needless listeners and missing the stated condition.
C. proposes NLB with path rules, which NLB cannot evaluate because it never parses HTTP URLs.
D. combines NLB with host rules and IP preservation, still lacking Layer 7 path logic and TLS offload semantics of ALB.
ANS-C01 exam tip
Path in the stem means ALB, host means ALB with host conditions, TLS plus header means HTTPS listener with X-Forwarded-For. NLB never does URL paths.
3A company has developed an application on AWS that will track inventory levels of vending machines and initiate the restocking process automatically. The company plans to integrate this application with vending machines and deploy the vending machines in several markets around the world. The application resides in a VPC in the us-east-1 Region. The application consists of an Amazon Elastic Container Service (Amazon ECS) cluster behind an Application Load Balancer (ALB). The communication from the vending machines to the application happens over HTTPS.
The company is planning to use an AWS Global Accelerator accelerator and configure static IP addresses of the accelerator in the vending machines for application endpoint access. The application must be accessible only through the accelerator and not through a direct connection over the internet to the ALB endpoint.
Which solution will meet these requirements?
Configure the ALB in a private subnet of the VPC. Attach an internet gateway without adding routes in the subnet route tables to point to the internet gateway. Configure the accelerator with endpoint groups that include the ALB endpoint. Configure the ALB’s security group to only allow inbound traffic from the internet on the ALB listener port.
Configure the ALB in a private subnet of the VPC. Configure the accelerator with endpoint groups that include the ALB endpoint. Configure the ALB's security group to only allow inbound traffic from the internet on the ALB listener port.
Configure the ALB in a public subnet of the VPAttach an internet gateway. Add routes in the subnet route tables to point to the internet gateway. Configure the accelerator with endpoint groups that include the ALB endpoint. Configure the ALB's security group to only allow inbound traffic from the accelerator's IP addresses on the ALB listener port.
Configure the ALB in a private subnet of the VPC. Attach an internet gateway. Add routes in the subnet route tables to point to the internet gateway. Configure the accelerator with endpoint groups that include the ALB endpoint. Configure the ALB's security group to only allow inbound traffic from the accelerator's IP addresses on the ALB listener port.
Answer: D
The short version
D — Private ALB locked to accelerator IPs with IGW routes gives accelerator-only access. The vending fleet uses static accelerator IPs and the ALB must block direct hits. Only a private placement plus security group allowlisting of accelerator addresses enforces that.
Key concepts in this question
Global Accelerator: anycast static IPs front accelerator, traffic rides AWS backbone to endpoint groups near the ALB.
ALB subnet posture: private subnets have no direct internet path, public subnets do, placement controls reachability.
Internet gateway routing: even private ALB needs IGW plus routes so return path and accelerator health checks complete correctly.
Security group allowlist: stateful filter on ALB listener that permits only accelerator IP ranges and drops direct browser attempts.
ECS behind ALB: containers receive only accelerator-sourced flows, preserving single entry point and audit simplicity.
Why D is correct
Option D places the ALB in a private subnet, attaches an IGW with proper routes for accelerator delivery, registers the ALB as accelerator endpoint, and restricts the ALB security group to accelerator source IPs on the listener port. That blocks anyone resolving the ALB DNS from hitting it directly while letting accelerator edge nodes forward vending HTTPS cleanly.
Why the others are wrong
A. allows inbound from the whole internet on the listener, so direct ALB access remains possible and violates isolation.
B. omits IGW and route entries, leaving accelerator to ALB delivery incomplete and health checks unreliable.
C. puts the ALB in a public subnet, making direct internet reach inherent even with SG tuning, weakening the only-through-accelerator goal.
ANS-C01 exam tip
Accelerator-only means private ALB plus SG set to accelerator IPs. If the SG says internet, it fails isolation. If subnets are public, it invites direct hits.
4A global delivery company is modernizing its fleet management system. The company has several business units. Each business unit designs and maintains applications that are hosted in its own AWS account in separate application VPCs in the same AWS Region. Each business unit's applications are designed to get data from a central shared services VPC.
The company wants the network connectivity architecture to provide granular security controls. The architecture also must be able to scale as more business units consume data from the central shared services VPC in the future.
Which solution will meet these requirements in the MOST secure manner?
Create a central transit gateway. Create a VPC attachment to each application VPC. Provide full mesh connectivity between all the VPCs by using the transit gateway.
Create VPC peering connections between the central shared services VPC and each application VPC in each business unit's AWS account.
Create VPC endpoint services powered by AWS PrivateLink in the central shared services VPCreate VPC endpoints in each application VPC.
Create a central transit VPC with a VPN appliance from AWS Marketplace. Create a VPN attachment from each VPC to the transit VPC. Provide full mesh connectivity among all the VPCs.
Answer: C
The short version
C — PrivateLink endpoint services in the center with endpoints in each spoke is most secure and scalable. Business units only need shared data, not mesh to each other. Only PrivateLink gives unidirectional, non-routed private access per service.
Key concepts in this question
PrivateLink: provider endpoint service backed by NLB, consumer interface endpoints with private IPs, no peering routes or internet.
Granular control: each service can be allowlisted per account, with security groups on endpoints limiting ports and callers.
Scale pattern: new business unit adds an endpoint, central side needs no route table churn or overlapping CIDR coordination.
Transit gateway mesh: full IP routing between all VPCs, broad blast radius unless heavily segmented with route tables.
Peering and VPN hub: peering is point to point and operationally heavy, VPN hub adds latency, cost and full mesh risk.
Why C is correct
Option C creates VPC endpoint services powered by PrivateLink in the shared services VPC and drops interface endpoints into each application VPC. Traffic stays on the private network, each unit reaches only the exposed service on approved ports, and no transitive routing lets spokes talk to each other.
Why the others are wrong
A. builds transit gateway full mesh, giving every spoke potential paths to every other spoke, opposite of least privilege.
B. uses peering per VPC, which scales as N connections, needs CIDR non-overlap and lacks service-level allowlisting.
D. builds VPN hub with marketplace appliance, adding encryption overhead, single choke and still full mesh reachability.
ANS-C01 exam tip
Most secure plus scale from center means PrivateLink. Full mesh language signals transit gateway trap, peering signals scale trap.
5A company uses a 4 Gbps AWS Direct Connect dedicated connection with a link aggregation group (LAG) bundle to connect to five VPCs that are deployed in the us-east-1 Region. Each VPC serves a different business unit and uses its own private VIF for connectivity to the on-premises environment. Users are reporting slowness when they access resources that are hosted on AWS.
A network engineer finds that there are sudden increases in throughput and that the Direct Connect connection becomes saturated at the same time for about an hour each business day. The company wants to know which business unit is causing the sudden increase in throughput. The network engineer must find out this information and implement a solution to resolve the problem.
Which solution will meet these requirements?
Review the Amazon CloudWatch metrics for VirtualInterfaceBpsEgress and VirtualInterfaceBpsIngress to determine which VIF is sending the highest throughput during the period in which slowness is observed. Create a new 10 Gbps dedicated connection. Shift traffic from the existing dedicated connection to the new dedicated connection.
Review the Amazon CloudWatch metrics for VirtualInterfaceBpsEgress and VirtualInterfaceBpsIngress to determine which VIF is sending the highest throughput during the period in which slowness is observed. Upgrade the bandwidth of the existing dedicated connection to 10 Gbps.
Review the Amazon CloudWatch metrics for ConnectionBpsIngress and ConnectionPpsEgress to determine which VIF is sending the highest throughput during the period in which slowness is observed. Upgrade the existing dedicated connection to a 5 Gbps hosted connection.
Review the Amazon CloudWatch metrics for ConnectionBpsIngress and ConnectionPpsEgress to determine which VIF is sending the highest throughput during the period in which slowness is observed. Create a new 10 Gbps dedicated connection. Shift traffic from the existing dedicated connection to the new dedicated connection.
Answer: A
The short version
A — Per-VIF throughput metrics identify the heavy business unit, then add a new 10 Gbps dedicated link. Each VPC has its own private VIF, so per-VIF bytes reveal the culprit. Only a new dedicated connection adds real capacity because existing dedicated bandwidth cannot be lifted in place.
Key concepts in this question
Private VIF metrics: VirtualInterfaceBpsEgress and Ingress in CloudWatch report per VIF throughput, mapping directly to one business unit VPC.
Connection metrics: ConnectionBps and Pps aggregate the whole physical link, hiding which VIF saturates the LAG bundle.
Direct Connect scaling: dedicated connections are fixed capacity physical ports, upgrade means provisioning a new higher speed port and migrating VIFs.
LAG saturation: shared bundle bursts at peak, so headroom must come from added ports.
Hosted versus dedicated: hosted connections top at lower speeds and suit smaller needs, not a saturated dedicated LAG fix.
Why A is correct
Option A queries VirtualInterfaceBpsEgress and Ingress around the slow hour to rank VIFs and pinpoint the surge. It then creates a new 10 Gbps dedicated connection and shifts traffic, relieving the 4 Gbps LAG with true added capacity.
Why the others are wrong
B. diagnoses correctly but tries to upgrade bandwidth of the existing dedicated port, which is not an allowed in-place operation.
C. uses connection-level metrics that cannot attribute load to a VIF, plus proposes a 5 Gbps hosted link that mismatches dedicated scaling.
D. pairs connection-level metrics with a new dedicated link, so capacity is right but attribution stays blind to the offending unit.
ANS-C01 exam tip
Per VIF equals VirtualInterface metrics, whole pipe equals Connection metrics. Dedicated speed goes up only via a new port, never an edit.
6A software-as-a-service (SaaS) provider hosts its solution on Amazon EC2 instances within a VPC in the AWS Cloud. All of the provider's customers also have their environments in the AWS Cloud.
A recent design meeting revealed that the customers have IP address overlap with the provider's AWS deployment. The customers have stated that they will not share their internal IP addresses and that they do not want to connect to the provider's SaaS service over the internet.
Which combination of steps is part of a solution that meets these requirements? (Choose two.)
Deploy the SaaS service endpoint behind a Network Load Balancer.
Configure an endpoint service, and grant the customers permission to create a connection to the endpoint service.
Deploy the SaaS service endpoint behind an Application Load Balancer.
Configure a VPC peering connection to the customer VPCs. Route traffic through NAT gateways.
Deploy an AWS Transit Gateway, and connect the SaaS VPC to it. Share the transit gateway with the customers. Configure routing on the transit gateway.
Answer: A, B
The short version
AB — NLB front plus PrivateLink endpoint service connects overlapping VPCs privately. Overlapping IPs rule out peering and transit, and internet is forbidden. Only PrivateLink with NLB bridges identical ranges without routing conflicts.
Key concepts in this question
Overlapping CIDRs: same IP ranges in provider and customer VPCs make peering and transit routing ambiguous and therefore unsupported.
PrivateLink model: provider exposes service via endpoint service, consumer creates interface endpoint with local private IPs, no route exchange.
NLB requirement: endpoint services must be fronted by Network Load Balancer, ALB cannot back an endpoint service.
SaaS scale: one service endpoint serves many customers without per-customer peering or shared transit tables.
Why AB is correct
Options A and B together place the EC2 SaaS fleet behind a Network Load Balancer, wrap it in an endpoint service, and grant each customer permission to create an interface endpoint in their own VPC. Consumers reach the service via private endpoint IPs in their own address space, so overlaps never collide and no traffic traverses the internet. Customer internals stay hidden, provider needs no knowledge of customer ranges, and onboarding is an approval plus endpoint creation.
Why the others are wrong
C. uses ALB for the endpoint service, which PrivateLink does not support as a service front.
D. uses peering plus NAT, which fails with overlapping IPs and still needs route coordination plus address disclosure.
E. uses shared transit gateway, which also requires non-overlapping CIDRs and creates broad transitive connectivity.
ANS-C01 exam tip
Overlap plus no internet equals PrivateLink with NLB. If you see peering or transit with overlap, eliminate. ALB never backs an endpoint service.
7A network engineer is designing the architecture for a healthcare company's workload that is moving to the AWS Cloud. All data to and from the on-premises environment must be encrypted in transit. All traffic also must be inspected in the cloud before the traffic is allowed to leave the cloud and travel to the on-premises environment or to the internet.
The company will expose components of the workload to the internet so that patients can reserve appointments. The architecture must secure these components and protect them against DDoS attacks. The architecture also must provide protection against financial liability for services that scale out during a DDoS event.
Which combination of steps should the network engineer take to meet all these requirements for the workload? (Choose three.)
Use Traffic Mirroring to copy all traffic to a fleet of traffic capture appliances.
Set up AWS WAF on all network components.
Configure an AWS Lambda function to create Deny rules in security groups to block malicious IP addresses.
Use AWS Direct Connect with MACsec support for connectivity to the cloud.
Use Gateway Load Balancers to insert third-party firewalls for inline traffic inspection.
Configure AWS Shield Advanced and ensure that it is configured on all public assets.
Answer: D, E, F
The short version
DEF — MACsec Direct Connect, GWLB firewall insertion and Shield Advanced cover encryption, inspection and DDoS. Healthcare needs on-prem encryption, inline cloud inspection before egress, plus internet protection with cost shield. Only that trio maps one to each mandate.
Key concepts in this question
MACsec on Direct Connect: Layer 2 encryption for dedicated links, satisfying all data to and from on-prem must be encrypted in transit.
Gateway Load Balancer: transparent insertion of third-party virtual appliances for inline inspection of north-south and east-west flows.
Shield Advanced: always-on detection plus mitigation on public assets with cost protection against scaling during attacks.
WAF scope: protects HTTP layer on selected resources, not a general inline network inspection fabric.
Mirroring versus inline: Traffic Mirroring copies for analysis but does not block, so it cannot enforce must be inspected before leaving.
Why DEF is correct
D supplies encrypted hybrid path via Direct Connect with MACsec, meeting the transit encryption rule without VPN overlays. E inserts firewalls via Gateway Load Balancer so traffic is inspected inline before heading to on-prem or internet. F places Shield Advanced on public appointment components for DDoS resilience and bill protection.
Why the others are wrong
A. offers Traffic Mirroring copy for forensics, not inline blocking, so uninspected traffic could still egress.
B. puts WAF everywhere, but WAF is Layer 7 filtering on specific resources, not full network inspection or link encryption.
C. uses Lambda to add security group denies, which is reactive automation, not guaranteed inline inspection or transit encryption.
ANS-C01 exam tip
Encrypted hybrid equals MACsec Direct Connect, must inspect before leave equals GWLB, DDoS plus bill protection equals Shield Advanced.
8A retail company is running its service on AWS. The company’s architecture includes Application Load Balancers (ALBs) in public subnets. The ALB target groups are configured to send traffic to backend Amazon EC2 instances in private subnets. These backend EC2 instances can call externally hosted services over the internet by using a NAT gateway.
The company has noticed in its billing that NAT gateway usage has increased significantly. A network engineer needs to find out the source of this increased usage.
Which options can the network engineer use to investigate the traffic through the NAT gateway? (Choose two.)
Enable VPC flow logs on the NAT gateway's elastic network interface. Publish the logs to a log group in Amazon CloudWatch Logs. Use CloudWatch Logs Insights to query and analyze the logs.
Enable NAT gateway access logs. Publish the logs to a log group in Amazon CloudWatch Logs. Use CloudWatch Logs Insights to query and analyze the logs.
Configure Traffic Mirroring on the NAT gateway's elastic network interface. Send the traffic to an additional EC2 instance. Use tools such as tcpdump and Wireshark to query and analyze the mirrored traffic.
Enable VPC flow logs on the NAT gateway's elastic network interface. Publish the logs to an Amazon S3 bucket. Create a custom table for the S3 bucket in Amazon Athena to describe the log structure. Use Athena to query and analyze the logs.
Enable NAT gateway access logs. Publish the logs to an Amazon S3 bucket. Create a custom table for the S3 bucket in Amazon Athena to describe the log structure. Use Athena to query and analyze the logs.
Answer: A, D
The short version
AD — VPC flow logs on the NAT ENI to CloudWatch or S3 with Insights or Athena reveal the heavy talkers. NAT gateways emit no native access logs, so flow telemetry on the elastic interface is the supported path. Both destinations enable IP, port and byte analysis.
Key concepts in this question
NAT gateway logging gap: no dedicated NAT access logs exist, flow logs on the gateway elastic network interface fill the gap.
Flow log fields: capture source, destination, ports, bytes, action and interface, enough to rank EC2 clients driving cost.
CloudWatch Insights path: stream ENI flow logs to a log group, then query top talkers interactively without managing storage.
S3 plus Athena path: archive ENI flow logs to S3, define table schema, run SQL for long-range or high-volume billing forensics.
Mirroring limits: Traffic Mirroring needs supported source types and capture fleet, not available on NAT gateway interfaces.
Why AD is correct
Options A and D both enable VPC flow logs scoped to the NAT gateway elastic network interface, capturing private-to-public translations that incur processing and data charges. Publishing to CloudWatch Logs allows Logs Insights queries for quick spike triage, while publishing to S3 with Athena supports durable, low-cost analytics over large windows. Using both gives fast and deep investigation without inventing a log type AWS does not provide.
Why the others are wrong
B. assumes NAT gateway access logs to CloudWatch, a log family that does not exist for managed NAT gateways.
C. tries Traffic Mirroring on the NAT interface with tcpdump, which is unsupported as a mirror source and adds capture overhead.
E. repeats the non-existent NAT access log idea to S3 with Athena, so destination is fine but source log type is invalid.
ANS-C01 exam tip
NAT cost spike means flow logs on the NAT ENI. Remember there is no NAT access log, and NAT cannot be a mirror source.
9A banking company is successfully operating its public mobile banking stack on AWS. The mobile banking stack is deployed in a VPC that includes private subnets and public subnets. The company is using IPv4 networking and has not deployed or supported IPv6 in the environment. The company has decided to adopt a third-party service provider's API and must integrate the API with the existing environment. The service provider’s API requires the use of IPv6.
A network engineer must turn on IPv6 connectivity for the existing workload that is deployed in a private subnet. The company does not want to permit IPv6 traffic from the public internet and mandates that the company's servers must initiate all IPv6 connectivity. The network engineer turns on IPv6 in the VPC and in the private subnets.
Which solution will meet these requirements?
Create an internet gateway and a NAT gateway in the VPC. Add a route to the existing subnet route tables to point IPv6 traffic to the NAT gateway.
Create an internet gateway and a NAT instance in the VPC. Add a route to the existing subnet route tables to point IPv6 traffic to the NAT instance.
Create an egress-only Internet gateway in the VPAdd a route to the existing subnet route tables to point IPv6 traffic to the egress-only internet gateway.
Create an egress-only internet gateway in the VPC. Configure a security group that denies all inbound traffic. Associate the security group with the egress-only internet gateway.
Answer: C
The short version
C — Egress-only internet gateway gives outbound-only IPv6 from private subnets. Servers must start all IPv6 flows and receive no inbound internet initiation. Only the egress-only gateway filters inbound while allowing stateful outbound.
Key concepts in this question
IPv6 egress-only: VPC component that permits outbound IPv6 plus return traffic but drops any internet-initiated inbound.
Private subnet routing: route tables point IPv6 default to the egress-only gateway, keeping instances unaddressable from outside.
NAT gateway scope: managed NAT handles IPv4 translation only, it does not process IPv6 flows.
Internet gateway behavior: full bidirectional IPv6 via IGW would expose private hosts to inbound internet, violating the mandate.
Security group myth: groups attach to instances and ENIs, never to gateway resources themselves.
Why C is correct
Option C creates an egress-only internet gateway and routes IPv6 traffic from the private subnets to it, so bank servers can call the third-party IPv6 API and receive responses while internet hosts cannot open new connections inward. That matches the company rule that servers initiate all IPv6 and the VPC stays IPv4 for existing workloads with added IPv6 only where needed. It avoids NAT constructs that cannot carry IPv6 and avoids IGW exposure that would permit inbound.
Why the others are wrong
A. points IPv6 at a NAT gateway, which does not forward IPv6, so API calls would fail.
B. points IPv6 at a NAT instance, which is an IPv4 oriented workaround and not the managed IPv6 egress path.
D. tries to attach a denying security group to the gateway itself, which is architecturally invalid and provides no routing.
10A company has deployed an AWS Network Firewall firewall into a VPC. A network engineer needs to implement a solution to deliver Network Firewall flow logs to the company’s Amazon OpenSearch Service (Amazon Elasticsearch Service) cluster in the shortest possible time.
Which solution will meet these requirements?
Create an Amazon S3 bucket. Create an AWS Lambda function to load logs into the Amazon OpenSearch Service (Amazon Elasticsearch Service) cluster. Enable Amazon Simple Notification Service (Amazon SNS) notifications on the S3 bucket to invoke the Lambda function. Configure flow logs for the firewall. Set the S3 bucket as the destination.
Create an Amazon Kinesis Data Firehose delivery stream that includes the Amazon OpenSearch Service (Amazon Elasticsearch Service) cluster as the destination. Configure flow logs for the firewall Set the Kinesis Data Firehose delivery stream as the destination for the Network Firewall flow logs.
Configure flow logs for the firewall. Set the Amazon OpenSearch Service (Amazon Elasticsearch Service) cluster as the destination for the Network Firewall flow logs.
Create an Amazon Kinesis data stream that includes the Amazon OpenSearch Service (Amazon Elasticsearch Service) cluster as the destination. Configure flow logs for the firewall. Set the Kinesis data stream as the destination for the Network Firewall flow logs.
Answer: B
The short version
B — Firehose to OpenSearch is the fastest managed path for firewall flow logs. Network Firewall can stream directly to Firehose, which batches and loads into OpenSearch near real time. That avoids S3 hop latency and custom polling.
Key concepts in this question
Firewall log destinations: Network Firewall flow logs support S3, CloudWatch Logs and Kinesis Data Firehose as configured targets.
Firehose delivery: managed buffering, transform and direct put into OpenSearch without operating consumers.
Shortest time goal: every extra store plus Lambda plus notification adds minutes, direct streaming minimizes delay.
Data stream gap: Kinesis Data Streams needs a consumer to land data, it is not a direct firewall log target for OpenSearch.
OpenSearch direct myth: firewall cannot select the search cluster itself as a log destination, it needs Firehose in front.
Why B is correct
Option B creates a Firehose delivery stream with the OpenSearch cluster as destination, then points Network Firewall flow logs at that stream. Logs flow from firewall to Firehose to indexed documents with managed retries and sub-minute latency, which is the documented low-latency pattern for search analytics. It removes S3 staging, SNS fan-out and Lambda maintenance while meeting the shortest possible time mandate with native AWS integration.
Why the others are wrong
A. stages in S3 then SNS to Lambda to OpenSearch, adding storage delay plus code to operate for no speed benefit.
C. sets OpenSearch as direct firewall destination, which the service does not offer, so configuration would fail.
D. uses Kinesis Data Streams as destination, which still needs a delivery consumer and is not the firewall to OpenSearch express path.
ANS-C01 exam tip
Firewall to search fastest equals Firehose to OpenSearch. S3 plus Lambda is slowest, direct cluster choice is invalid.