Intro to HPC - Archive
- Simon Alexander
- Lisa Stafford
Introduction: This workshop will introduce the core concepts behind High Performance Computing and the various services that ARCC provides. After the workshop, participants will understand:
The various services that ARCC provides.
The basic infrastructure that makes up a cluster.
Course Goals:
To introduce ARCC’s (HPC Center) Mission and Services.
What are Clusters: ARCC and NWSC.
What is HPC?
What does a HPC/Cluster Architecture look like.
Different types of storage.
01: Introduction to HPC
Topics:
ARCC (HPC Center) Mission and Services.
Clusters: ARCC and NWSC.
What is HPC?
HPC/Cluster Architecture.
Different types of storage.
ARCC
NWSC
The NCAR Wyoming Supercomputing Center (NWSC) represents a collaboration between NCAR and the University of Wyoming.
NCAR
NCAR: National Center for Atmospheric Research:
A center of research excellence in Earth system science sponsored by the National Science Foundation.
World-Class Research in Earth System Science
CISL: Computational & Information System Lab
ARCC: Mission
Provide support for research computing endeavors including:
high performance computing
large research data storage
and consulting
to further the University of Wyoming and State of Wyoming’s strategic priorities by enabling researchers with the necessary computational needs.
Core Service 1: High Performance Computing: HPC
We maintain a number of clusters for the purpose of allowing researchers to perform a variety of use cases such as running:
Computation-intensive analysis on large datasets.
Long large-scale simulations.
10s/100s/1000s of small short tasks - nothing is too small.
and lots of other use case…
Core Service 2: Research Data Storage
Safe and secure storage and transfer of data that researchers can share and collaborate on with others within UW, and other institutions across the world.
High performance data storage geared toward project-oriented data.
Storage for published research data.
Low-cost storage solution that enables a Cloud-like presence for research data hosted by ARCC.
Hosting onsite backups and enabling data sharing and collaboration.
Core Service 2: Research Data Storage: Which One?
Considerations: Cost vs Usability:
Consider as more traditional storage that can be accessed via SMB/AD via a traditional Windows File Explorer/Globus.
Access follows the idea of a project that users are part of and authenticated via username/AD.
A cheaper storage solution that is accessed either via a client and/or programmatically that uses S3 to provide object storage via buckets.
Access it provides is via access/secret key tokens, that can be time based.
Data can be made publicly available.
It does not user the notion of projects/usernames.
Come and discuss what your needs and use cases are…
Core Service 2: Research Data Storage Changes:
Effective June 1, 2024, ARCC introduced the ‘ARCC Data Portal’ serving the dual purpose of providing high performance back end storage for the MedicineBow HPC system and a data storage solution for researchers needing a centralized data repository for ongoing research projects.
Data Portal storage is FREE up to the default allocation quota.
ARCC’s Data Portal is compromised of VAST data storage compromised of high speed all-NVMe storage, housing 3 petabytes of raw storage. VAST storage employs data de-duplication allowing the system to logically store more than the raw 3PB available.
Alcova storage on the ARCC Data Portal can be thought of as the “new Alcova” and will replace the prior Alcova storage space listed here. This space is intended for use as collaborative data storage space using SMB protocol for interactive access. This space is backed up by ARCC and can only be used by researchers with a uwyo.edu account.
MedBow space can be thought of as the root level directory of the HPC system, separated into home, project, and gscratch directories, intended for use with HPC workflows where speed and minimal overhead are prioritized over backups.
– the essence of these services will remain
– but the underlying systems will be updating.
Core Service 3: End User Support
Available to all researchers at UW: User Services
Website/Wiki - examples and suggested best practices.
Zoom office hours - Hosted twice a week
Tuesdays 11am-1pm
Wednesdays 12-2pm
One-on-one consultation.
Scheduled in-person and (online) trainings.
YouTube Channel.
Other Services:
SouthPass: Web based access to the clusters - included Jupyter Notebooks.
Linux Desktop Support.
Hosting of services - R Shiny application…
A GitLab service.
Proposal Development.
As research needs grow, so will services that we offer.
We’re always open to constructive feedback and suggestions.
We’re here to provide the services for you…
HPC: High Performance Computing
“High Performance Computing most generally refers to the practice of aggregating computing power in a way that delivers much higher performance than one could get out of a typical desktop computer or workstation in order to solve large problems in science, engineering, or business.”
HPC ≠ Desktop
HPC >> Desktop
What is a Cluster:
Compute Nodes:
Core Service 1: HPC: What does this look like?
We maintain a number of clusters for the purpose of allowing researchers to perform a variety of use cases such as running:
Computation-intensive analysis on large datasets.
Megabytes / Gigabytes / Terabytes.
On the filesystem in one / many files.
In memory.
CPU only vs GPU enabled.
Long large-scale simulations.
Hours, days, weeks…
Single job across multiple nodes each using multiple cores.
10s/100s/1000s of small short tasks - nothing is too small.
Seconds, minutes, hours…
Single node - one to many cores.
and lots of other use case…
UW IT Data Center:
Types of HPC systems:
There are generally two type of HPC systems:
Homogenous: All compute nodes in the system share the same architecture. CPU, memory, and storage are the same across the system.
Heterogenous: The compute nodes in the system can vary architecturally with respect to CPU, memory, even storage, and whether they have GPUs or not.
Typically, similar compute nodes are grouped via partitions.
Cluster and Partitions:
Condominium Model:
The “condo model”.
Allow researchers to invest into the cluster - purchasing additional compute nodes that they get priority to use.
‘preempt’ jobs outside of the investor’s project - allow the investor to start their jobs immediately.
“immediately” if no other jobs from that investment project are already using the investment.
A preempted job is stopped and automatically re-queued. When it starts will be determined by the current cluster utilization.
Consider the idea of check-pointing which allows a job to continue analysis at the point where it was stopped.
This is managed by defining ‘investor partitions’.
Ask for Assistance:
Zoom Office Hours.
When requesting help:
Be as clear and specific as you can.
Provide enough detail so we can replicate your issue.
Provide job ids, log files, working folder paths…
Software and module versions.
Links to your software homepage/repo and what you’ve tried.
Have you consulted online communities?
When using other people’s code: try and understand what it is doing before asking for help.
We’ll always do our best effort, but we’re not domain experts (supporting all researchers).
Next Steps: Request an Account with ARCC
Summary
Covered:
ARCC (HPC Center) Mission and Services.
Clusters: ARCC and NWSC.
What is HPC?
HPC/Cluster Architecture.
Different types of storage.