PEARC 2025 Cultivating Computational Equity Central Valley

Citation

Sarvani Chadalapaka, Jeffrey Weekley, Emily Jane McTavish, Anna Klimaszewski-Patterson and Ruth Marinshaw. “Cultivating Computational Equity: Building a Collaborative Research Computing Community in California’s Central Valley.” In Practice and Experience in Advanced Research Computing (PEARC ‘25), July 20-24, 2025, Columbus, OH. ACM, 4 pages. https://doi.org/10.1145/3708035.3736054. Copy: raw/shared/2026-10-28 PEARC 2025 Cultivating Computational Equity Central Valley.md (original: .pdf beside it). Read 2026-10-28.

Bottom line

• Four Central Valley campuses used almost no national ACCESS compute in 2024. Each had 0.02 to 0.04 percent of CPU allocations, and CSU Stanislaus had none (§2).
• The paper says the gap is people and skills, not only hardware. It argues for “intermediate clusters that offer both computational resources and researcher support” (§2).
• CENVAL-ARC is a California model for a UC campus that leads training and shared compute with nearby California State University (CSU) campuses.

Background

CENVAL-ARC is the Central Valley Accessible Research and Computational Hub.

The second author, Jeffrey Weekley, is Director, Research IT, UC Santa Cruz.

The other authors are Emily Jane McTavish (UC Merced), Anna Klimaszewski-Patterson (CSU Sacramento) and Ruth Marinshaw (Yale University).

Key claims

• Members: “University of California Merced (UCM) and California State University (CSU) campuses in Fresno, Sacramento, and Stanislaus” (§2).
• ACCESS use in 2024, as a share of total ACCESS CPU allocations (§2):
  • CSU Fresno: “A 0.04% allocation, 3 PI projects, and 6 users”.
  • CSU Sacramento: “A 0.03% allocation, 2 projects, and 4 users”.
  • UC Merced: “A 0.02% allocation, with 13 faculty members accessing resources”.
  • CSU Stanislaus: “No allocation, no project, and no users”.
• Cost barrier, from a participant: “I wanted to scale beyond my local campus resources to cloud resources like Amazon, but who’s paying for Amazon? No one.” (§1)
• Challenges found (§2.1):
  • “Limited Local Expertise”: too few research computing facilitators and HPC system administrators.
  • “Insufficient Computational Training”: gaps in command line, parallel computing, machine learning and data-intensive methods.
  • “Access Barriers to National Resources”: national centers cannot reach every campus. Regional groups have “geographical proximity, local trust”.
• Student body, Fall 2022 (Table 1):

• “UCM is the only doctorate-granting institution in California’s Central Valley.” All sites are “within a two-hour drive of UC Merced” (§2.2).
• Scale argument: small and mid-scale clusters “share many features with largescale HPC”. So “building skills locally is a key step” toward national resources (§2).
• Training: an online “Introduction to Unix” session in February 2025, “attended by over 30 participants from multiple campuses” (§3).
• Symposium: a multi-day event at UC Merced on HPC, high-throughput computing (HTC) and data-intensive workflows (§4). Objectives included the Campus Research Computing Consortium (CaRCC) capabilities model and use of Open Science Grid (OSG) systems (§4.1).
• Outreach: “With no prior regional event of this kind”, organizers used personal email campaigns and faculty networks (§4.2).
• Plans: quarterly meetings, training in cloud computing, data management and software practice, and richer participant data (§5).

Evidence and method

A four-page practice paper in ACM proceedings. The ACCESS figures are a 2024 snapshot. The demographic table cites outside data for Fall 2022. Feedback is a few quoted comments, not a survey. The authors state that they used OpenAI’s Consensus as a research tool and ChatGPT-4 for some headings (Acknowledgments).

Limits

• The paper gives no node counts, GPU numbers, cost or access rules for the CENVAL-ARC cluster.
• It gives no symposium attendance total. Figure 3 shows roles, but the copy has no figure data.
• The ACCESS share is CPU only. It does not cover GPU allocations or other national programs.
• The paper is by the program’s own team.