HPC in the Cloud

High performance cloud computing powered by Amazon Web Services

In addition to our on-premises HPC clusters, ACER provides access to a full-featured, on-demand computational cluster hosted on the AWS Linux 2 Cloud. It is capable of running a wide variety of CPU and GPU-accelerated workloads. Researchers are charged on a monthly basis determined by the total active CPU and disc usage accumulated during the billing cycle.

The cluster has a large number of software packages installed on its computers including CUDA and  MPI libraries, C and C++ compilers, and the python and R interpreted languages. If you find a package is not available on the cluster, please feel free to place a request to have it added.

Please click here to submit a request for an account on our AWS HPC Cloud Cluster.

Available Hardware Configurations

Hardware configuration of a GPU compute node on the ACER Cloud Cluster.

Prices displayed below may not be up to date. Please visit AWS On-Demand Pricing and use the AWS Instance Names listed below to find the current rates (select US East (Ohio) for the region and Linux for the OS).

  • gpu-g4-t4
    • 1 x NVIDIA Tesla T4 GPU; 16 GiB GPU memory
    • 2 x Intel Xeon Scalable Processors @ 2.5 GHz
    • 4 x vCPU
    • 16 GiB RAM
    • AWS Instance Name: g4dn.xlarge
    • $0.526 USD/hour
  • gpu-p3-v100
    • 1 x NVIDIA Volta V100 GPU; 16 GiB GPU memory
    • 4 x Intel Xeon Scalable Processors @ 2.7 GHz
    • 8 x vCPU
    • 61 GiB RAM
    • AWS Instance Name: p3.2xlarge
    • $3.06 USD/hour

Hardware configuration of a CPU compute node on the HPC Cloud Cluster.

Prices displayed below may not be up to date. Please visit AWS On-Demand Pricing and use the AWS Instance Names listed below to find the current rates (select US East (Ohio) for the region and Linux for the OS).

  • cpu-t2
    • 2 x Intel Xeon Scalable Processors @ 2.3 GHz
    • 2 x vCPU
    • 8 GiB RAM
    • AWS Instance Name: t2.large
    • $0.0928 USD/hour
  • cpu-t3
    • 2 x Intel Xeon Scalable Processors @ 2.5 GHz
    • 4 x vCPU
    • 16 GiB RAM
    • AWS Instance Name: t3.xlarge
    • $0.1664 USD/hour
  • cpu-m4
    • 8 x Intel Xeon Scalable Processors @ 2.3 GHz
    • 16 x vCPU
    • 64 GiB RAM
    • AWS Instance Name: m4.4xlarge
    • $0.8 USD/hour
  • cpu-c5
    • 36 x Intel Xeon Scalable Processors @ 3.4 GHz
    • 72 x vCPU
    • 192 GiB RAM
    • AWS Instance Name: c5n.18xlarge
    • $3.888 USD/hour
  • highmem-x1e
    • 32 x Intel Xeon Scalable Processors @ 2.3 GHz
    • 64 x vCPU
    • 1952 GiB RAM
    • AWS Instance Name: x1e.16xlarge
    • $13.344 USD/hour

The HPC Cloud cluster user home directories are located in /shared/home. This directory has a quota limit of 50 GB per user. Also available is a non-persistent high performance Lustre filesystem, /lustre. This filesystem should be used for jobs requiring high i/o. Users need be aware that because the Lustre filesystem is non-persistent, any output written to /lustre that needs to be saved should be copied to the /shared/home directory once the job is finished.

Quick Links Heading link