Sizing summary and recommendations
|
RStudio Professional |
Minimum |
Recommended |
|
RStudio Workbench (previously RStudio Server Pro) |
2 core / 4G (RAM) / 100 G (Disk) |
4 - 16 core / 8G - 256G (RAM) / |
|
RStudio Connect |
2 core / 4G (RAM) / |
8 - 16 core / 32G - 128G (RAM) / |
|
RStudio Package Manager |
2 core / 2G (RAM) / 200 G (Disk) |
2 core / 4 G (RAM) / 500 G (Disk) |
Recommendations
RStudio Professional products run on Linux servers and require R to be installed first. By default, R is a single threaded process that holds all its data in memory. Our products have relatively small footprints, therefore the minimum requirements for running RStudio products depends mostly on R. A minimum instance with 2 cores and 4G of memory is enough to support multiple concurrent R processes as well as our products.
We recommend more cores and more memory for best use of our products. The main drivers that determine server size include: number and size of concurrent R sessions. The number of R developers and end users can help determine the number of concurrent R sessions. The amount of data your organization uses with R will determine the size of those sessions.
For example, 2-3 data scientists who with a million records and build small applications will probably require a small server. A team of 15 data scientists who work with big data and deliver results to dozens of end users will require a large server.
These recommendations are guidelines only. You may require more resources if your requirements are greater. If you are using virtual machines that have memory caps, you may require multiple instances. Our products are designed to scale across multiple instances.
RStudio Workbench
Hardware Configuration
Local Execution
The recommended setup is a production cluster with two or more RStudio Workbench instances. The instances can be clustered together with a built-in load balancer. Note: an RStudio cluster balances user sessions as opposed to R jobs. A cluster setup requires a shared home mount. Users can access their data via ODBC connections or shared file servers. Also recommended is a test server to help with administering the environment.
External Execution
Beginning with version 1.2, RStudio Server Pro (now RStudio Workbench) can be configured to offload code execution to resource managers such as Kubernetes or Slurm. In this configuration, the computational and memory requirements of RStudio Workbench can be fairly minimal, since the execution takes place in a separate environment. In this case, the hardware sizing recommendations below can be used to guide decisions about the external resource requirements for individual R sessions.
Hardware Sizing
The most important drivers for sizing the environment are (1) the number of concurrent R sessions; and (2) the average sessions size. R is memory intensive, so it’s best to get as much RAM as possible. If you use virtual machines you might have restrictions on how much memory you can allocate to a single instance. In that case we recommend getting as much memory as possible and consider using multiple nodes.
- Minimum (2 core / 4G). This server will be for testing and sandboxing.
- Small (4 core / 8G). This server will support one or two analysts with tiny data.
- Large (16 core / 256G). This server will support 15 analysts with a blend of session sizes. Alternatively, it will support dozens of analysts with very small sessions.
- Other. Larger instances (e.g. 1 TB RAM) may be useful for heavier workloads.
Scratch space for RStudio Workbench is highly variable. A lot of scratch work can be pushed to file shares and databases, as opposed to local disk. Generally speaking, R will not need as much scratch space as SAS since R does not write its intermediate data files to disk. Nevertheless, it's always good to have some scratch space on the R server. Depending on data volumes and the number of users, 250 Gb - 1 Tb of disk space per server will be sufficient for most use cases. If you plan to support large data workflows, then we recommend more than 1 Tb of disk space.
In order for Project Sharing to work, the storage volumes containing the project directories must support POSIX Access Control Lists (ACLs). ACLs are used to grant collaborators access to shared projects without changing other file permissions.
RStudio Connect
Hardware Configuration
RStudio Connect can be installed as a single server deployment or setup in a load balanced, highly available cluster. RStudio Connect is designed to host all of the work your team publishes with R. That includes documents, apps, APIs, and plots. In addition to hosting, RStudio Connect can also schedule reports and automate simple workflows that depend on R. RStudio Connect can be configured to send out emails on a regular schedule on behalf of users.
Hardware Sizing
RStudio Connect's hardware specifications will depend on the number and type of applications, documents, and analysis running on the server. Standard specifications for a production server might range from 8-16 cores and 32-128 GB of RAM. Some workloads may require substantially larger servers. We highly recommend consulting with the R programmers to determine the resources required to run the Shiny applications and automated R Markdown processes.
- Minimum (2 core / 4G). This server is good for testing.
- Small (8 core / 32G). This server will support plenty of lightweight Shiny apps, plumber APIs, and simple R Markdown documents. For example, you might have a small team of data scientists who build small apps and post static documents for 20 end users.
- Large (16 core / 128G). This server will support large apps and complex R artifacts. For example, you might have a large team of data scientists who build large apps and schedule complex documents for 100 end users.
- Cluster Setting up 2 or more nodes allows for job distribution across the cluster and high availability. Cluster configurations require a PostgreSQL database and shared storage.
- A minimum of 1GB of storage is recommended for the PostgreSQL database server.
RStudio Package Manager
Hardware Configuration
The recommended setup is a high availability cluster with two RStudio Package Manager instances. Follow the instructions in the admin guide for configuring a cluster. We also recommend using a staging server to test changes and configurations.
Hardware Sizing
Unlike other RStudio products, RStudio Package Manager focuses on storing and organizing packages not running computations. For that reason, disk space is the most important consideration for RStudio Package Manager as opposed to CPU and RAM.
- Minimum (2 core / 2G of RAM / 200 GB disk). This server is useful for testing and staging.
- Recommended (2 servers, each with 2 cores, 4 GB of RAM, 500 GB disk). This cluster can support multiple repositories and can be used by a team or even multiple departments. See the admin guide for details on storage requirements, as they may vary in air-gapped environments or in cases where multiple operating systems and R versions are required.
Getting started
Where to download R
You must install R before installing any RStudio product. RStudio does not ship or support R. RStudio products will work with any version of R including, CRAN, Microsoft, Oracle, etc.
Linux required
RStudio Professional products run the linux operating system.
45 day evaluation
All RStudio Professional products offer a 45 day evaluation.
Learn more
All RStudio Professional products support enterprise authentication and security. They are designed to scale with your organization. And they are supported by our team. Each product has a host of additional features to make these tools enterprise ready. For more information see the products online.
Comments