Examkin

Question

Difficulty: Very hardHigh-Performing and Elastic Compute Solutions

An engineering firm is designing a high-performance compute (HPC) system on AWS to run tightly-coupled computational fluid dynamics (CFD) simulations. The simulations utilize Message Passing Interface (MPI) for inter-node communication and require sub-millisecond network latency and high throughput. The control interface of the worker nodes runs on a custom service bound to TCP port 8443. An Application Load Balancer (ALB) will distribute control traffic to these worker nodes. The compute cluster must dynamically scale based on demand. Which architecture satisfies these requirements while maximizing performance and ensuring high availability of the control traffic?

  1. A
    Deploy the worker nodes on Amazon EC2 instances across a spread placement group, enabling Elastic Fabric Adapter (EFA) on the network interfaces. Register the instances to an ALB target group with the health check port explicitly configured to 8443.
  2. B
    Deploy the worker nodes on Amazon EC2 instances within a cluster placement group, enabling Elastic Fabric Adapter (EFA) on the network interfaces. Register the instances to an ALB target group, leaving the health check port at the default configuration to simplify administration.
  3. Deploy the worker nodes on Amazon EC2 instances within a cluster placement group, enabling Elastic Fabric Adapter (EFA) on the network interfaces. Register the instances to an ALB target group with the health check port explicitly configured to 8443.Answer
  4. D
    Deploy the simulation workloads using AWS Lambda functions configured with a 15-minute timeout and provisioned concurrency. Route control traffic via an ALB with target groups pointing to the Lambda functions on port 8443.

Answer

Deploy the worker nodes on Amazon EC2 instances within a cluster placement group, enabling Elastic Fabric Adapter (EFA) on the network interfaces. Register the instances to an ALB target group with the health check port explicitly configured to 8443.
The correct architecture uses a cluster placement group to achieve low-latency node-to-node communication, which is necessary for tightly-coupled MPI applications. Elastic Fabric Adapter (EFA) is enabled to provide OS-bypass for MPI, which significantly improves latency and throughput. Finally, the Application Load Balancer target group's health check port must be explicitly configured to 8443 because the control interface is bound to that custom TCP port. This prevents the health checks from failing, which would otherwise happen if left on the default port 80.

Step-by-Step Solution

1
Analyze the networking latency requirements of the HPC workload.
Tightly-coupled fluid dynamics simulations using MPI require sub-millisecond node-to-node latency.
This requirement determines the selection of placement groups and network adapters.
2
Choose the optimal EC2 placement group and network adapter configuration.
Select a cluster placement group to group instances close together, and enable Elastic Fabric Adapter (EFA).
Cluster placement groups minimize physical network distance. EFA provides OS-bypass capabilities, bypassing the operating system kernel to accelerate MPI communications.
3
Evaluate the Application Load Balancer health check configuration for the custom application port.
Configure the target group's health check port to explicitly use 8443.
The control traffic on the worker nodes listens on port 8443. Leaving the health check at default would query port 80, causing health checks to fail and making the backend instances unavailable.

Key Concept

High-Performance Compute Clustering and Load Balancer Health Check Ports
Rate this question

Topics

High-Performing and Elastic Compute Solutions

More questions from this topic