Red Cloud Computing & GPUs
Red Cloud is a subscription-based Infrastructure as a Service (IaaS) cloud platform that provides on-demand virtual servers and storage with root access. It offers a cost-effective alternative to public cloud services, with no oversubscription of compute resources.
Cornell-Exclusive Offer
Cornell students, faculty, and staff are eligible for a free Red Cloud Exploratory Project Account, which includes:
- 165 core hours
- 50GB of storage
- 1 hour of consulting assistance
Quick Links
Red Cloud Compute
- Launch virtual machines with up to 128 CPU cores and 1TB RAM in seconds.
- All CPU cores and RAM are dedicated, not shared or oversubscribed — ensuring fast and consistent performance.
- You're only charged while your instances are running.
- Choose from a variety of configurations that fit the needs of your application, including NVIDIA T4 and V100 GPUs.
- Supported operating systems: Rocky Linux, Ubuntu Linux, and Windows.
Storage
Powered by a Ceph cluster with 1.9 petabytes of raw capacity, Red Cloud offers:
- Volumes – persistent disk storage that can be attached/detached from instances.
- Object Storage – fully Amazon S3-compatible; access via any S3 client/SDK or through the Red Cloud Object Storage collection on Globus.
- Distributed File System – mount CephFS on Linux instances for reliable, high-performance storage.
- Your first subscription includes 50GB of storage. Additional storage can be purchased as needed.
Note: Red Cloud is not suitable for Cornell high-risk or restricted data, or for state or federally
regulated data. Data stored on Red Cloud is not backed up by default; users are responsible for their own backups.
Cohesity is a solution available to Cornell users.
Network
- 40 Gigabit connectivity and direct connection to Cornell's 100 Gigabit Internet2 link.
- Manage access by IP address and port for strong security controls.
- No additional charges for network traffic.
User Interface
- Built on OpenStack.
- Manage your environment using the Red Cloud web console, OpenStack command line clients, or any development libraries supporting the OpenStack API.
Applications
Virtual HPC Cluster with Slurm Scheduler
MATLAB Parallel Server
- MATLAB Parallel Server users must be familiar with the MATLAB Parallel Computing Toolbox (PCT).
- PCT is included in Cornell's site license.
- PCT tutorial available or request consulting assistance when starting your project.
- Ensure installation of PCT-compatible clients on your local machine.
- Review supported MATLAB clients and startup requirements.
More Software
Subscription Model
- 1 Red Cloud subscription = 8,585 core hours
- That's equivalent to 1 core running 24/7 for 1 year
- Use core hours flexibly—consume them fast or slow depending on your compute needs
- Think of it like a prepaid Starbucks card—no surprise bills
- Discounts are available for multiple subscriptions
Need Help?