High Performance Computing — HPC
Shared high performance computing resources for Columbia researchers.
Also known as HPC, Shared HPC, and SRCPAC HPC.
CUIT’s High Performance Computing service provides a cluster of computing resources that power transactions across numerous research groups and departments at the University, as well as additional projects and initiatives as demand and resources allow. The Shared Research Computing Policy Advisory Committee (SRCPAC) oversees the operation of existing HPC clusters through faculty-led subcommittees.
HPC service is available 24x7, however downtimes for maintenance may be scheduled every 3 months. The duration of these planned outages varies but is typically less than a day and is announced to users in advance via email and CUIT Service Alerts.
Upcoming High Performance Computing Trainings
- February 11 | UC San Diego: HPC Performance Tuning and Optimization
- This webinar is open to all higher-ed attendees as part of an NSF grant. This session will cover the effective use of cache, loop-level optimizations, force reductions, optimizing compilers and their limitations, short-circuiting, time-space tradeoffs and more.
- March 26 | Using Google Cloud at Columbia: Fundamentals for Researchers
-
This session aims to cover how to create a Columbia-linked GCP account (and the benefits of doing so), how to set up an instance (server) in the cloud for data analysis, how to run a job / scripts, how to access LionMail Drive data in GCP, how to store data in the cloud.
-
Classic HPC Workshop Recordings
- Accessing “ACCESS”, a Free NSF-Funded HPC Resource [January 27, 2025 recording]
- Using HPC to Enhance Your Research: A Series of Three Classes
- Using Jupyter Notebooks on HPC [April 10, 2024 recording]
Current High Performance Computing Clusters
Insomnia Shared HPC Cluster
Insomnia went live in February 2024 and initially was a joint purchase by 21 research groups and departments. Unlike its predecessors, this new high performance computing design allows researchers to buy not only a node but half or even a quarter of a node on the cluster.
Insomnia is faculty-governed by the cross-disciplinary SRCPAC and is administered and supported by CUIT’s High Performance Computing team.
Insomnia is a new type of design intended to allow indefinite expansion to Columbia's shared high performance computing cluster, adding new hardware and capabilities as needed. It is a perpetual cluster. Hardware will be retired after five (5) years.
43 nodes with a total of 6880 cores (80 physical cores per node, doubled via hyperthreading)
All of Insomnia's high performance computing servers are equipped with Dual Intel Xeon Platinum 8460Y processors (2 GHz):
- Standard Nodes (512Gi): 28 nodes
- GPU Nodes: 7 nodes
- High Memory Nodes (1.0Ti): 8 nodes
- Total combined CPUs: 6,880 CPUs
- Total combined RAM: 23.05 TiB
- Total combined GPUs: 14 GPUs
- 291TB GPFS filesystem
- HDR Infiniband
- Red Hat Enterprise Linux 9.3
- Slurm job scheduler
Ginsburg Shared HPC Cluster
Ginsburg high performance computing went live in February 2021 and is a joint purchase by 33 research groups and departments.
The cluster is faculty-governed by the cross-disciplinary SRCPAC and is administered and supported by CUIT’s High Performance Computing team.
Tentative retirement dates
- Ginsburg Phase 1 retirement: February 2026
- Ginsburg Phase 2 retirement: March 2027
- Ginsburg Phase 3 retirement: December 2027
286 nodes with a total of 9152 cores (32 cores per node):
All of Ginburg's high performance computing servers are equipped with Dual Intel Xeon Gold 6226R processors (2.9 GHz):
- 191 Standard Nodes (192 GB)
- 56 High Memory Nodes (768 GB)
- 18 RTX 8000 GPU nodes (2 GPUs modules per server)
- 4 V100S GPU nodes (2 GPU modules per server)
- 8 A100 GPU Nodes (2 GPU modules per server)
- 9 A40 GPU Nodes (2 GPU modules per server)
- 1PB of DDN ES7790 Lustre storage
- HDR Infiniband
- Red Hat Enterprise Linux 8
- Slurm job scheduler
Terremoto Shared HPC Cluster
The Terremoto high performance computing cluster was launched in December 2018, and is located in the Columbia University Data Center.
The cluster is faculty-governed by the cross-disciplinary SRCPAC and is administered and supported by CUIT’s High Performance Computing team.
Tentative retirement dates
- Terremoto Phase 1 retirement: Completed December 2023
- Terremoto Phase 2 retirement: Q1 2025
137 nodes with a total of 3288 cores (24 cores per node)
Dell C6420 nodes with dual Intel Xeon Gold 6126 Processor (2.6 Ghz):
- 111 Standard Nodes (192 GB)
- 14 High Memory Nodes (768 GB)
- 12 GPU Nodes with two Nvidia V100 GPU modules
- EDR Infiniband
- Red Hat Enterprise Linux 7
- Slurm job scheduler
Retired High Performance Computing Clusters
The Habanero high performance computing cluster, retired in December 2023, was located in the Shared Research Computing Facility (SRCF), a dedicated portion of the university data center on the Morningside campus.
Yeti, the high performance computing cluster retired in 2019, was located in the Shared Research Computing Facility (SRCF), a dedicated portion of the university data center on the Morningside campus.
Hotfoot, now retired, was launched as a high performance computing cluster in 2009 as a partnership among: the departments of Astronomy & Astrophysics, Statistics, and Economics plus other groups represented in the Social Science Computing Committee (SSCC); the Stockwell Laboratory; CUIT; and the Office of the Executive Vice President for Research; and Arts & Sciences.
In later years the cluster ran the Torque/Moab resource manager/scheduler software and consisted of 32 nodes which provided 384 cores for running jobs. The system also included a 72 TB array of scratch storage.
CUIT offers four ways to leverage the computing power of our High Performance Computing resources.
Researchers may review a server "menu" and pricing and submit an order via our HPC Cluster Purchase page . A variety of high performance computing purchasing options are available with pricing tiers that reflect the level of computing capability purchased. Purchasers receive higher priority than others users.
An individual researcher may pay a set fee for a share of the high performance computing system for one year as a single user with the ability to use additional computing capacity as it is available, based on system policies and availability. The current price is set at $1000/year.
Submit a request form for an HPC rental request now.
Researchers, including graduate students, post-docs, and sponsored undergraduates may use the high performance computing system on a low-priority, as-available basis. User support is limited to online documentation only.
Submit a request form for free HPC access now.
Instructors teaching a course or workshop addressing an aspect of computational research may request temporary access to a high performance computing cluster for their students. Access will typically be arranged in conjunction with a class project or assignment.
Submit a request form for HPC Education access now.
Current high performance computing contacts can request access to their HPC group for a new user by emailing [email protected]. This option is available to current authorized contacts only.
Other Columbia Research Resources
The following organizations are available to support research initiatives at Columbia:
- Executive Vice President for Research (EVPR): Office of Research Initiatives
- Columbia Libraries: Digital Scholarship
Computing Resources Outside of Columbia
CUIT's Research Computing Services team has put together a list of additional, fee-based, third-party high performance computing and storage platforms as a resource for our research community.
At this time, RCS does not recommend one resource over another, or provide support for these external services.
- ACCESS HPC (from NSF)
- Amazon Web Services including Elastic Compute Cloud (EC2) and storage services
- Columbia University Center for Computational Biology and Bioinformatics (C2B2) colocation and hosting services
- Cornell University Center for Advanced Computing and Red Cloud services
- National Artificial Intelligence Research Resource (NAIRR) Pilot
- New York State HPC Program: Resources at RPI, Brookhaven
- New York State HPC Consortium: Web-based resources at RPI, Brookhaven, Stony Brook, and University of Buffalo
- San Diego Supercomputing Center (SDSC) Cloud Storage Services
- USC Digital Repository
- Windows Azure