Learn about EC2 and HPC (High Performance Computing) in the below section.

  • AWS provides elastic and scalable infrastructure for HPC applications.
  • integrated services to build and manage HPC clusters in AWS
  • Handle compute intensive workloads for
    • genomics
    • computational chemistry
    • financial risk modeling
    • computer aided engineering
    • weather prediction
    • seismic imaging
    • machine learning
    • deep learning
    • autonomous driving

AWS HPC Working

EC2 and HPC (High Performance Computing)
HPC Components
EC2 and HPC (High Performance Computing)

Data Management & Data Transfer

  • Snowball and Snowmobile transfer large amounts of data into and out of the AWS.
  • DataSync, automate moving data between on-premises storage and S3 or EFS
  • Direct Connect establish private connectivity between AWS and datacenter, office, or colocation environment

Compute

  • choose from variety of compute instance types AS PER needs
  • They include
    • Intel® Xeon® processor-powered CPU instances – C5n, C5d and Z1d
    • GPU-based instances
    • field programmable gate array (FPGA)-powered instances.
  • C5n instances feature Intel Xeon Platinum 8000 series (Skylake-SP) processor provides
    • sustained all core Turbo CPU
    • clock speed of up to 3.5 GHz.
    • up to 100 Gbps of network bandwidth
    • up to 14 Gbps of dedicated bandwidth to Amazon EBS.
  • For high-speed, ultra-low latency local storage workload requirement use C5d instances
    • equipped with local NVMe-based SSDs
  • EC2 z1d instances offer both high compute capacity and high memory footprint
    • deliver sustained all core frequency of up to 4.0 GHz
  • EC2 P3dn instances based on GPU, feature
    • 100 Gbps network bandwidth (up to 4x the bandwidth of previous P3 instances)
    • local NVMe storage
    • latest NVIDIA V100 Tensor Core GPUs with 32 GB of GPU memory
    • NVIDIA NVLink for faster GPU-to-GPU communication
  • Auto Scaling monitors applications and automatically adjusts capacity
  • Auto Scaling easy to setup application scaling for multiple resources across multiple services in minutes

Networking

  • EC2 instances support enhanced networking to
  • achieve higher bandwidth
  • lower inter-instance latency
  • Elastic Fabric Adapter (EFA) a network interface for EC2 instances enables running HPC applications
    • Provides high levels of inter-node communications
    • Has custom-built operating system
    • bypass hardware interface for high performance in inter-instance communications
  • placement groups for tightly-coupled HPC applications needing low latency networking.
  • Virtual Private Cloud provides IP connectivity between compute instances and storage components.

Storage

  • Multiple storage options as per performance and costs are provided
  • AWS offers flexible object, block, or file storage
  • Amazon EBS gives persistent block storage volumes with EC2
  • Provisioned IOPS can allocate storage volumes as per need and attach to EC2 instances
  • S3 store any type of data over Internet and store HPC input and output data for long term
  • FSx for Lustre is a high performance file storage service for HPC workloads and
    • transparently presents S3 objects as files to write results back to S3
    • Use as standalone high-performance file system to burst workloads from on-premises to AWS
  • Amazon EFS provides simple, scalable file storage

Automation and Orchestration

  • Automating job submission and scheduling makes efficient use of HPC infrastructure.
  • AWS Batch can
  • run multiple batch computing jobs
  • dynamically provisions right type and quantity of compute resources as per job needs.
  • AWS ParallelCluster is
    • fully supported and maintained open source cluster management tool
    • makes it easy to deploy and manage HPC clusters in AWS
  • NICE EnginFrame is
    • web portal provides access to HPC-enabled infrastructure using web browser.
    • EnginFrame gives a user-friendly HPC
      • job submission
      • job control
      • job monitoring environment

Operations & Management

  • Amazon CloudWatch is
    • a monitoring and management service
    • provides data and actionable insights to monitor applications
    • understand and respond to system-wide performance changes
    • optimize resource utilization
    • get a unified view of operational health
  • AWS Budgets can
    • set custom budgets
    • alert you when costs or usage exceed budgeted amount.

Visualization Tools

  • Remote visualization accelerate turnaround times for engineering design
  • NICE DCV enables remote access 2D/3D interactive applications over a standard network.
  • Amazon AppStream 2.0 is managed application streaming service that securely deliver application sessions to web browser.

Security and Compliance

  • Multiple security related services and templates offered in AWS
  • strong safeguards to protect customer privacy
  • AWS Identity and Access Management (IAM) provides
    • a solution for managing users, roles, and groups
    • give rights to access specific data sources
    • issue users and systems individual identities and credentials
    • provision temporary access credentials using Amazon STS
  • Multiple compliance programs are complied, as
    • HIPAA
    • FISMA
    • FedRAMP
    • PCI
    • ISO 27001
    •  SOC 1, and others

Elastic Fabric Adapter (EFA)

  • EFA gives tightly-coupled HPC applications
    • lower and more consistent latency
    • higher throughput than traditional TCP channels
    • can be enabled dynamically, on-demand
  • Useful for Applications using MPI or supporting ibfabric communication stack
  • EFA devices provide all ENA devices functionalities plus a new OS bypass hardware interface
  • Communication using OS bypass functionality is with instances within a single subnet of VPC, only.
  • Currently available on C5n.18xlarge, and P3dn.24xl, and I3en.24xl instance sizes.
  • An ENA ENI provides IP networking features to support VPC networking
  • An EFA ENI gives
    • all features of ENA ENI
    • plus, hardware support for applications to communicate directly with the EFA ENI
    • without involving the instance kernel (OS-bypass communication)
  • EFA ENIs can only be attached at launch or to stopped instances.

NICE DCV

  • It is a graphics-optimized streaming protocol
  • suited for streaming productivity applications on mobile devices ,HPC simulation visualization, etc
  • On server side, supports Windows and Linux. And, on client side, supports Windows, Linux, and MacOS
  • Also supports Web Client for HTML5 browser based access across devices, on client side
  • native clients support more features as multi-monitor support
  • the Windows native client also support USB support for 3D mice, storage devices and smart cards.
  • Also offer NICE DCV with 3D graphics-intensive applications needing low latency
  • Apt for HPC applications as
    • seismic and reservoir simulations
    • computational fluid dynamics (CFD) analyses
    • 3D molecular modeling
    • VFX compositing
    •  Game Engine based 3D rendering
  • Is a custom protocol to transfer images generated on server to client and conversely
  • allows client to control server’s keyboard and mouse
  • leverages WebSocket and TLS for security and compatibility
  • NICE DCV support all Amazon EC2 x86-64 architecture based instance types
  • No, need for a license server to install and use NICE DCV server on an EC2 instance.
  • NICE DCV is a downloadable software, can be downloaded and installed on running sessions.
  • By default, NICE DCV server is configured to communicate over port 8443.
  • can specify a custom TCP port after installing NICE DCV server and port must be greater than 1024.
Menu