Quando uma instância EC2 é criada uma interface de rede(ENI - Elastic Network Interface) também é criada e associada a subnet da instância.
As ENIs, quando criadas, recebem um endereço IPv4 privado e em determinadas condições também recebem um endereço IPv4 público.
É possível conectar múltiplos adaptadores de rede a uma instância, os adaptadores podem residir em diferentes subnets mas as subnets devem estar na mesma Availability Zone.
An elastic network interface is a logical networking component in a VPC that represents a virtual network card. It can include the following attributes:
- A primary private IPv4 address from the IPv4 address range of your VPC
- One or more secondary private IPv4 addresses from the IPv4 address range of your VPC
- One Elastic IP address (IPv4) per private IPv4 address
- One public IPv4 address
- One or more IPv6 addresses
- One or more security groups
- A MAC address
- A source/destination check flag
- A description
You can create a network interface, attach it to an instance, detach it from an instance, and attach it to another instance. The attributes of a network interface follow it as it's attached or detached from an instance and reattached to another instance. When you move a network interface from one instance to another, network traffic is redirected to the new instance.
Each instance has a default network interface, called the primary network interface. You cannot detach a primary network interface from an instance. You can create and attach additional network interfaces. The maximum number of network interfaces that you can use varies by instance type.
In a VPC, all subnets have a modifiable attribute that determines whether network interfaces created in that subnet (and therefore instances launched into that subnet) are assigned a public IPv4 address. he public IPv4 address is assigned from Amazon's pool of public IPv4 addresses. When you launch an instance, the IP address is assigned to the primary network interface that's created.
When you create a network interface, it inherits the public IPv4 addressing attribute from the subnet. If you later modify the public IPv4 addressing attribute of the subnet, the network interface keeps the setting that was in effect when it was created. If you launch an instance and specify an existing network interface as the primary network interface, the public IPv4 address attribute is determined by this network interface.
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-eni.html#AvailableIpPerENI
ENA is a custom network interface optimized to deliver high throughput and packet per second (PPS) performance, and consistently low latencies on EC2 instances. Using ENA, customers can utilize up to 20 Gbps of network bandwidth on certain EC2 instance types. ENA-based Enhanced Networking is currently supported on X1 instances, and will be available on other new EC2 instance types in the future. https://aws.amazon.com/pt/about-aws/whats-new/2016/06/introducing-elastic-network-adapter-ena-the-next-generation-network-interface-for-ec2-instances/
You can enable enhanced networking on supported instance types to provide lower latencies, lower network jitter, and higher packet-per-second (PPS) performance. Most applications do not consistently need a high level of network performance, but can benefit from access to increased bandwidth when they send or receive data. https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/accelerated-computing-instances.html
Elastic Fabric Adapter (EFA) is a network interface for Amazon EC2 instances that enables customers to run applications requiring high levels of inter-node communications at scale on AWS.
With EFA, High Performance Computing (HPC) applications using the Message Passing Interface (MPI) and Machine Learning (ML) applications using NVIDIA Collective Communications Library (NCCL) can scale to thousands of CPUs or GPUs. As a result, you get the application performance of on-premises HPC clusters with the on-demand elasticity and flexibility of the AWS cloud.
Use Cases:
- Computational Fluid Dynamics
- Weather modeling
- Machine Learning
https://aws.amazon.com/pt/hpc/efa/
An EFA is an Elastic Network Adapter (ENA) with added capabilities. It provides all of the functionality of an ENA, with an additional OS-bypass functionality. OS-bypass is an access model that allows HPC and machine learning applications to communicate directly with the network interface hardware to provide low-latency, reliable transport functionality.
With an EFA, HPC applications use MPI(Message Passing Interface) or NCCL(NVIDIA Collective Communication Library) to interface with the Libfabric API. The Libfabric API bypasses the operating system kernel and communicates directly with the EFA device to put packets on the network. This reduces overhead and enables the HPC application to run more efficiently.
https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa.html
