HPSee HPSee
down

HPSee Requirements

Find Your HPSee Setup

Your setup

1. Where do you want to run HPSee?
On premises
On AWS
AWS

2. How is your on-premise hardware organized?
Machines dedicated to HPSee
A cluster controlled by a scheduling system, such as SLURM
Small or large cluster
run with Apptainer or a hybrid of Docker and Apptainer

3. How many machines do you dedicate to HPSee?
One machine
Single machine
Run with Docker Compose or Apptainer
Several machines
Small or large cluster
run with Docker Swarm or Apptainer

Your Setup: AWS

It looks like this is the perfect solution for you. The following section provides detailed technical information about the required hardware, primarily intended for the relevant department and IT administrator.

AWS

Run HPSee as a managed cluster on AWS, with Amazon Aurora PostgreSQL as the database. Worker requirements apply per job, as in the other setups.

Setup at a glance

Worker nodes

Up to 300

Managed cluster on AWS

Manager node

16 vCPU / 32 GB

Plus the database

Database

Amazon Aurora

PostgreSQL

Worker, per job

8 vCPU / 16 GB

Recommended RAM; minimum 8 GB

Deployment and layout

How HPSee runs Managed cluster
Worker nodes Up to 300
Database Amazon Aurora PostgreSQL
Manager node 16 vCPU and 32 GB RAM, plus the database

Database

AWS uses Amazon Aurora PostgreSQL. The local database-host and container memory-budget values in the on-premises sections do not specify an Aurora configuration. Consult the deployment guide for the AWS installation.

Service requirements

The manager-node specification above describes the setup allocation. The API values below describe the service contribution and should not be added again if already covered by that allocation.

API

CPU
2 vCPU
Recommended: 4 vCPU
RAM
4 GB
Recommended: 8 GB
Disk (/datastorage)
At least 20 GB
Recommended: 100 GB+
Network
1 GbE
Recommended: 10 GbE if users download whole libraries

Worker, per job

CPU 8 vCPU
RAM
8 GB
Recommended: 16 GB
Disk (/datastorage)
1 GB per concurrent job, at least 10 GB
Recommended: 20 GB NVMe
Network
1 GbE
Recommended: 10 GbE at 100+ nodes
Running several jobs at once?

Multiply worker CPU and RAM by the number of concurrent jobs. Network requirements stay the same. For disk space, allow 1 GB per concurrent job, with at least 10 GB in total; 20 GB NVMe is recommended.

Server deployment

Use the AWS deployment described in the HPSee Deployment Guide. The Linux and Windows installation requirements in the on-premises sections apply to installations you run on your own machines.

SeeSAR client

The SeeSAR client connects to HPSee. Its requirements are listed separately from the server sizing above.

Memory (RAM) 16 GB, required for working with C-S-D results in SeeSAR
Operating systems Fully supported: Linux, macOS and Windows
Deployment guidance

All hardware requirements are intended as guidance. Actual needs depend on your workload, for example the number of jobs that run at the same time. Setups on dedicated machines need hardware reserved for HPSee, because HPSee does not throttle itself for other jobs on the same machine. The deployment options available to you depend on your HPSee version. The HPSee Deployment Guide describes the installation for each setup. We refine these requirements continuously as HPSee evolves and as we gain experience from more deployments, so values may be updated over time.

Back to setup overview ↑

Select Your Cluster Size

Small Cluster Large Cluster

Your Setup: Small Cluster

It looks like this is the perfect solution for you. The following section provides detailed technical information about the required hardware, primarily intended for the relevant department and IT administrator.

Small cluster

Run HPSee across worker nodes with a dedicated manager node. The database and API run on the manager; Slurm setups also run one EventBroker replica there. Only the number of workers changes with the size of the setup.

Setup at a glance

Worker nodes

10–20

Worker requirements apply to each concurrent job

Manager allocation

8 vCPU / 16 GB

Plus the database-host requirements below

Database RAM

64 GB

Recommended database-host contribution on the manager

Network

10 GbE

Recommended for the small-cluster database

Deployment and layout

Orchestration Docker Swarm or Slurm, with a dedicated manager node
Runtime Docker Swarm or Apptainer on dedicated machines; Apptainer or a hybrid of Docker and Apptainer for scheduling systems such as Slurm
Worker nodes 10–20
Database On the manager node
Manager allocation 8 vCPU and 16 GB RAM, plus the database-host requirements below
API One replica on the manager node
EventBroker One replica on the manager node in Slurm setups
Sizing the manager node

Add the database-host CPU and RAM contributions below to the manager allocation above. The API and, for Slurm setups, EventBroker service values describe their individual requirements; avoid counting those services again if they are already covered by your manager allocation. Workers are sized separately on the worker nodes.

Database host

CPU
8 vCPU
Recommended: 16 vCPU
Host RAM
32 GB
Recommended: 64 GB
Container memory budget*
24 GB
Recommended: 48 GB
Disk
1 TB NVMe, ≥ 500 MB/s, ≥ 5,000 IOPS
Recommended: 2 TB, ≥ 1 GB/s
Network
1 GbE
Recommended: 10 GbE
Database container memory budget

This budget is included in the database host RAM and must not be added again. * From HPSee 3.0, the installer asks for the container memory budget.

Service requirements

Each row lists the minimum first, with the recommended value beneath it where available.

API

CPU
2 vCPU
Recommended: 4 vCPU
RAM
4 GB
Recommended: 8 GB
Disk (/datastorage)
At least 20 GB
Recommended: 100 GB+
Network
1 GbE
Recommended: 10 GbE if users download whole libraries

EventBroker — Slurm setups only

Runs as one replica on the manager node in Slurm setups.

CPU 1 vCPU
RAM
512 MB
Recommended: 1 GB
Disk (/datastorage)
Image only, about 1 GB
Recommended: 5 GB with log headroom
Network Shared

Worker, per job

CPU 8 vCPU
RAM
8 GB
Recommended: 16 GB
Disk (/datastorage)
1 GB per concurrent job, at least 10 GB
Recommended: 20 GB NVMe
Network
1 GbE
Recommended: 10 GbE at 100+ nodes
Running several jobs at once?

Multiply worker CPU and RAM by the number of concurrent jobs. Network requirements stay the same. For disk space, allow 1 GB per concurrent job, with at least 10 GB in total; 20 GB NVMe is recommended.

HPSee server: operating systems

Linux
Supported
Docker Engine or Apptainer, depending on your setup
Windows
Supported
Windows 10 (version 1903, build 18362) or later, with WSL2 and Docker Desktop
macOS
Not currently supported
HPSee server installations are not currently supported on macOS

SeeSAR client

The SeeSAR client connects to HPSee. Its requirements are listed separately from the server sizing above.

Memory (RAM) 16 GB, required for working with C-S-D results in SeeSAR
Operating systems Fully supported: Linux, macOS and Windows
Deployment guidance

All hardware requirements are intended as guidance. Actual needs depend on your workload, for example the number of jobs that run at the same time. Setups on dedicated machines need hardware reserved for HPSee, because HPSee does not throttle itself for other jobs on the same machine. The deployment options available to you depend on your HPSee version. The HPSee Deployment Guide describes the installation for each setup. We refine these requirements continuously as HPSee evolves and as we gain experience from more deployments, so values may be updated over time.

Back to setup overview ↑

Your Setup: LargeCluster

It looks like this is the perfect solution for you. The following section provides detailed technical information about the required hardware, primarily intended for the relevant department and IT administrator.

Large cluster

Run HPSee across worker nodes with a dedicated manager node. The database and API run on the manager; Slurm setups also run one EventBroker replica there. Only the number of workers changes with the size of the setup.

Setup at a glance

Worker nodes

150

Worker requirements apply to each concurrent job

Manager allocation

16 vCPU / 32 GB

Plus the database-host requirements below

Database RAM

128–192 GB

Recommended database-host contribution on the manager

Network

10 GbE

Required for the large-cluster database

Deployment and layout

Orchestration Docker Swarm or Slurm, with a dedicated manager node
Runtime Docker Swarm or Apptainer on dedicated machines; Apptainer or a hybrid of Docker and Apptainer for scheduling systems such as Slurm
Worker nodes 150
Database On the manager node
Manager allocation 16 vCPU and 32 GB RAM, plus the database-host requirements below
API One replica on the manager node
EventBroker One replica on the manager node in Slurm setups
Sizing the manager node

Add the database-host CPU and RAM contributions below to the manager allocation above. The API and, for Slurm setups, EventBroker service values describe their individual requirements; avoid counting those services again if they are already covered by your manager allocation. Workers are sized separately on the worker nodes.

Database host

CPU
16 vCPU
Recommended: 32 vCPU
Host RAM
96 GB
Recommended: 128–192 GB
Container memory budget*
80 GB
Recommended: 128 GB
Disk
2 TB NVMe, ≥ 1 GB/s, ≥ 20,000 IOPS
Recommended: 4 TB, separate WAL (write-ahead log) volume
Network 10 GbE, required
Database container memory budget

This budget is included in the database host RAM and must not be added again. * From HPSee 3.0, the installer asks for the container memory budget.

Service requirements

Each row lists the minimum first, with the recommended value beneath it where available.

API

CPU
2 vCPU
Recommended: 4 vCPU
RAM
4 GB
Recommended: 8 GB
Disk (/datastorage)
At least 20 GB
Recommended: 100 GB+
Network
1 GbE
Recommended: 10 GbE if users download whole libraries

EventBroker — Slurm setups only

Runs as one replica on the manager node in Slurm setups.

CPU 1 vCPU
RAM
512 MB
Recommended: 1 GB
Disk (/datastorage)
Image only, about 1 GB
Recommended: 5 GB with log headroom
Network Shared

Worker, per job

CPU 8 vCPU
RAM
8 GB
Recommended: 16 GB
Disk (/datastorage)
1 GB per concurrent job, at least 10 GB
Recommended: 20 GB NVMe
Network
1 GbE
Recommended: 10 GbE at 100+ nodes
Running several jobs at once?

Multiply worker CPU and RAM by the number of concurrent jobs. Network requirements stay the same. For disk space, allow 1 GB per concurrent job, with at least 10 GB in total; 20 GB NVMe is recommended.

HPSee server: operating systems

Linux
Supported
Docker Engine or Apptainer, depending on your setup
Windows
Supported
Windows 10 (version 1903, build 18362) or later, with WSL2 and Docker Desktop
macOS
Not currently supported
HPSee server installations are not currently supported on macOS

SeeSAR client

The SeeSAR client connects to HPSee. Its requirements are listed separately from the server sizing above.

Memory (RAM) 16 GB, required for working with C-S-D results in SeeSAR
Operating systems Fully supported: Linux, macOS and Windows
Deployment guidance

All hardware requirements are intended as guidance. Actual needs depend on your workload, for example the number of jobs that run at the same time. Setups on dedicated machines need hardware reserved for HPSee, because HPSee does not throttle itself for other jobs on the same machine. The deployment options available to you depend on your HPSee version. The HPSee Deployment Guide describes the installation for each setup. We refine these requirements continuously as HPSee evolves and as we gain experience from more deployments, so values may be updated over time.

Back to setup overview ↑

Your Setup: Single Machine

It looks like this is the perfect solution for you. The following section provides detailed technical information about the required hardware, primarily intended for the relevant department and IT administrator.

Single machine

Run HPSee on one dedicated machine, with the database, API and worker sharing the same hardware. No separate manager node is needed. Deploy with Docker Compose or Apptainer, depending on your HPSee version.

Recommended sizing at a glance

Example for Docker Compose with one job, using the recommended service values where available. CPU, RAM and disk contributions add up because all services run on the same machine.

CPU

20+ vCPU

Database: 8+ cores
API: 4 vCPU · Worker: 8 vCPU

RAM

56 GB

Database: 32 GB
API: 8 GB · Worker: 16 GB

Disk

620 GB+

Database: 500 GB+ NVMe
API: 100 GB+ · Worker: 20 GB NVMe

Network

1 GbE

10 GbE if users download whole libraries.
The database uses local loopback.

Sizing notes

The database CPU recommendation is stated in cores; API and worker CPU values are stated in vCPU. The 20+ vCPU total follows the supplied sizing example. The 16 GB database container memory budget is already included in the 56 GB RAM total and must not be added again. For each extra concurrent job, add 8 vCPU and 16 GB RAM when using the recommended worker values.

Deployment and layout

Hardware One machine used only by HPSee
Deployment Docker Compose or Apptainer, depending on your HPSee version
Services Database, API and at least one worker, all on the same machine
Worker nodes 1
Database Container on the same machine
Manager node No separate manager node

Service requirements

Each row lists the minimum first, with the recommended value beneath it where available. These are individual service contributions; use the combined sizing example above to plan the whole machine.

Database

CPU
8 vCPU
Recommended: 8+ cores
Host RAM contribution
16 GB
Recommended: 32 GB
Container memory budget*
8 GB
Recommended: 16 GB
Disk
250 GB NVMe
Recommended: 500 GB+
Network Not applicable (loopback)
Database memory budget

This budget applies to the database only and is included in its host RAM contribution. It is not an additional RAM requirement. * From HPSee 3.0, the installer asks for the container memory budget.

API

CPU
2 vCPU
Recommended: 4 vCPU
RAM
4 GB
Recommended: 8 GB
Disk (/datastorage)
At least 20 GB
Recommended: 100 GB+
Network
1 GbE
Recommended: 10 GbE if users download whole libraries

Worker, per job

CPU 8 vCPU
RAM
8 GB
Recommended: 16 GB
Disk (/datastorage)
1 GB per concurrent job, at least 10 GB
Recommended: 20 GB NVMe
Network 1 GbE
Running several jobs at once?

Multiply worker CPU and RAM by the number of concurrent jobs. Network requirements stay the same. For disk space, allow 1 GB per concurrent job, with at least 10 GB in total; 20 GB NVMe is recommended.

HPSee server: operating systems

Linux
Supported
Docker Engine or Apptainer, depending on your setup
Windows
Supported
Windows 10 (version 1903, build 18362) or later, with WSL2 and Docker Desktop
macOS
Not currently supported
HPSee server installations are not currently supported on macOS

SeeSAR client

The SeeSAR client connects to HPSee. Its requirements are listed separately from the server sizing above.

Memory (RAM) 16 GB, required for working with C-S-D results in SeeSAR
Operating systems Fully supported: Linux, macOS and Windows
Deployment guidance

All hardware requirements are intended as guidance. Actual needs depend on your workload, for example the number of jobs that run at the same time. Setups on dedicated machines need hardware reserved for HPSee, because HPSee does not throttle itself for other jobs on the same machine. The deployment options available to you depend on your HPSee version. The HPSee Deployment Guide describes the installation for each setup. We refine these requirements continuously as HPSee evolves and as we gain experience from more deployments, so values may be updated over time.

Back to setup overview ↑