Training a model, fine-tuning one, or running a large analysis takes GPUs. As more courses teach machine learning, more students need them, and the campus platform has not always had enough for all of them. Some students have paid for cloud GPU time themselves to finish coursework and projects.
PARASOL is the campus response. It is the platform students already know, formerly called DSMLP, with a reservation system that matches capacity to the academic calendar, more GPUs from Winter 2027, and a name that reflects who it serves.
What the Name Stands For¶
PARASOL is short for Platform for Applied Research, Analytics, Science and Open Learning. The platform supports classroom GPU demand first, and runs research on the same cluster, with the same tools and support. From Winter 2027, Open Learning gives students in eligible majors and programs a workspace for coursework and exploration outside a course.
GPU Servers Arriving in Winter 2027¶
The Halıcıoğlu School of Data Science and Computing is adding four GPU servers to the platform in Winter 2027, nearly tripling its overall GPU capacity. The addition was enabled by Taner Halıcıoğlu's founding gift.
- 4GPU serversArriving in Winter 2027
- 32GPUs in totalNVIDIA RTX PRO 6000 Blackwell
- 66,800Peak GPU-hours a yearApproximate, for coursework and student projects
- 213,000+Off-peak GPU-hours a yearFor approved research, capstones, and graduate work
Courses, projects and research groups across campus can use the new servers. The campus is grateful to the School for funding them. The School has first claim on them, for its courses in peak weeks and for courses, research and approved projects at other times. An hour the School has not booked opens to the rest of the campus gradually, over the 12 hours before it starts.
GPU Hardware & CUDA lists the GPU models behind each GPU class, including the RTX PRO 6000 Blackwell that the new servers carry.
Why Reservations¶
GPU demand is uneven. It peaks around assignment deadlines and in peak weeks, weeks 6 to 10 of a quarter and finals week, and it falls overnight and on weekends. A reservation holds a GPU for a specific window, so a student with a deadline does not have to wait for a free GPU. Off-peak hours cost fewer Service Units, which moves long research runs to the hours when classes use the fewest GPUs.
Reservations explains how to book a window. Service Units & Budgets explains how GPU time is metered.
What Comes Next¶
The new servers will also give the campus better data on GPU demand. University leaders will use that data to plan further investment, including software and services that make GPU computing easier to use in courses outside science and engineering.
Industry Tools and Workflows¶
PARASOL runs containers on Kubernetes, with the same tools and workflows used in industry and research computing. Students in undergraduate, master's, and doctoral programs learn on the kind of infrastructure they will use after graduation.
See Get Started or How PARASOL Works.