Different types of network interfaces on AWS

What is an Elastic Network Interface (ENI) on AWS, and what are some of its key features?

An Elastic Network Interface (ENI) is a virtual network interface that you can attach to an instance in a VPC. Key features include a primary private IP address, one or more secondary private IP addresses, one Elastic IP address per private IP address, one public IP address (if the instance is in a public subnet), one or more security groups, a MAC address, and source/destination checking attribute. ENIs allow you to create a network topology within a VPC that is similar to a traditional network.

What is the purpose of the source/destination check attribute on an ENI, and how can it be modified?

The source/destination check attribute ensures that the ENI only sends and receives traffic that is destined for the instance to which it is attached. This check can be modified to “disabled” for instances that serve as a NAT instance, network appliance, or virtual firewall, allowing them to handle traffic not meant for them. This setting can be modified using the AWS Management Console, CLI, or API.

Can you explain the differences between an ENI and an Elastic Network Adapter (ENA)?

ENAs are designed to provide higher performance compared to ENIs. They support Enhanced Networking on larger instance types, offer higher bandwidth, high packet per second (PPS) performance, and consistently low latency. ENAs are most often used with newer instance types and are ideal for workloads requiring higher network performance.

How does an ENA contribute to Enhanced Networking in AWS, and what instances support it?

ENA enables Enhanced Networking which is a feature that uses a hardware-accelerated network interface to provide high-performance networking capabilities on supported EC2 instances. Enhanced Networking reduces CPU utilization for networking tasks and provides higher bandwidth, higher packet per second (PPS) rates, and lower inter-instance latencies. ENA supports C5, M5, R5, and other latest generation instance families.

What is an Elastic Fabric Adapter (EFA), and for what kind of applications is it best suited?

EFA is a network device that a customer can attach to their Amazon EC2 instance to accelerate High-Performance Computing (HPC) and Machine Learning applications. It supports low-latency, high-throughput inter-node communication at scale and is ideal for applications that require message-passing interface (MPI) capabilities.

In what ways can AWS Transit Gateway be considered a network interface, and what benefits does it provide?

While AWS Transit Gateway itself is not strictly a network interface, it acts as a regional virtual router for traffic flowing between VPCs, VPN connections, and AWS Direct Connect gateways. It simplifies network architecture, reduces operational overhead, and provides a hub-and-spoke model that centralizes and manages connectivity.

How is a Gateway Load Balancer Endpoint (GWLBE) different from a standard ENI, and when would you use it?

GWLBE is a type of elastic network interface that allows you to privately access services running on a Gateway Load Balancer from any VPC in the same region. It differs from a standard ENI in that it is specially designed to integrate with third-party virtual appliances such as firewalls and intrusion prevention systems. GWLBE simplifies the management of incoming traffic and provides easy scalability and high availability for network services.

Discuss the differences between an Instance Level Public IP and an Elastic IP on AWS.

An Instance Level Public IP (PIP) is automatically assigned to an instance and is directly tied to the instance’s lifecycle, meaning it is lost if the instance is stopped or terminated. In contrast, an Elastic IP (EIP) is a static, public IPv4 address that you can allocate to your account and associate with an instance or ENI. Unlike a PIP, an EIP remains associated with the account until it’s explicitly released, providing a reliable, fixed IP address for services.

How can Network Load Balancers (NLB) be used in relation to network interfaces, and what type of traffic are they best suited for?

Network Load Balancers operate at the fourth layer of the OSI model (transport layer) and are used to distribute TCP, UDP, and TLS traffic across multiple instances by targeting the instances’ ENIs. They’re best suited for balancing load with ultra-low latencies, extremely high volumes of sudden traffic, or stable traffic distribution among instances in a VPC.

What is the function of a VPC Endpoint in AWS networking architecture, and what interface does it provide?

A VPC Endpoint enables private connections between your VPC and supported AWS services or VPC endpoint services without requiring an Internet gateway, NAT device, VPN, or AWS Direct Connect connection. It provides a service-specific interface to interact with AWS services entirely within the AWS network, therefore increasing the security and reducing the latency of the traffic.

Can you describe a situation where using a Secondary Elastic IP (EIP) might be advantageous, and how would it be utilized with an ENI?

A Secondary EIP can be advantageous when hosting multiple websites on a single EC2 instance, requiring separate IP addresses for SSL certificates, or managing network traffic with distinct IP addresses. You would associate the Secondary EIP with a secondary private IP address tied to the ENI of the EC2 instance, then configure the operating system and server software to listen on the secondary IP address.

Explain how the concept of IP address affinity is relevant to network interfaces like ELBs or NLBs.

IP address affinity, also known as session stickiness, is relevant to ELBs and NLBs as it ensures that requests from a particular client are directed to the same target behind the load balancer. This is important for maintaining session state between the client and the server, which can be critical for applications that rely on user session information. In the case of NLBs, this is achieved through TCP connections, while ELBs also offer application-level stickiness.