Recently, Starfish received a pleasant surprise—in the form of a support ticket—from Todd Green, who had taken over as Director of IT at the Scientific Computing and Imaging Institute (SCI) at the University of Utah.
SCI’s mission is to tackle problems of societal importance through collaborative, translational computing research in imaging, simulation, visualization, data science, scientific and high-performance computing, and more. It’s also home to the One-U Responsible Artificial Intelligence Initiative, which is growing the university’s pool of AI researchers and expanding cyberinfrastructure to support interdisciplinary AI research with real-world impact. (Fun fact: SCI is pronounced “ski,” a nod to Utah’s Olympic slopes.)
The support ticket Todd sent was about a surprising discovery he made while using Starfish—and how it quickly turned a potential data disaster into an impressive win.
Starfish migrates 2.3 billion files without a hiccup
SCI is a research powerhouse, where massive datasets are the norm. When it came time to replace an inherited, undocumented aging 2016 storage solution with a modern data platform, Todd’s team faced a monumental task:
- Migrate 2.3 billion files
- Change the UIDs on all files to comply with campus-wide authentication and authorization for data sharing
- Avoid downtime, to not impact grant deadlines and ongoing research
- Ensure all files were properly transferred and the copied data matched the source, through Starfish’s hashing and comparison features
After hearing a glowing review from a colleague at Los Alamos National Lab and meeting the Starfish team at Supercomputing 2023, SCI began their Starfish evaluation. It quickly became clear: Starfish was the right fit.
As Todd put it:
“We can’t have downtime. We have grant deadlines and paper submissions. This means our compute servers run 24/7. The only way I could see doing this was with Starfish.”
The initial migration went off without a hitch. Using Starfish’s background sync and parallel agents, SCI seamlessly moved data to the new two-petabyte server. Many users didn’t even notice.
A window left open – researchers were still putting data in old locations
However, months later, something odd happened. Turns out, the old storage system had an additional, undocumented access path that was still being written to for a single dataset.
“A researcher said, ‘Hey, the storage is near capacity. I don’t understand.’ After some digging, I responded, ‘You’re writing to the old area… How are you writing to the old storage? I changed that to be read-only!’”
With thousands of undocumented exports that had been set up by the previous team who managed the system, it wasn’t readily evident that there were multiple paths to the same dataset. It was like locking the front door, only to find users had been slipping in through an open side window.
Starfish closes the window and fixes the problem in under an hour
With researchers writing data to both the old and new locations, syncing them without overwriting recent changes seemed like a nightmare.
But not with Starfish.
“Changing the extra export to read-only, two quick Starfish scans, a treediff, and then some copies and other minor tweaking… and in under an hour I had everything trued back up.”
Todd was able to:
- Instantly identify differences between directories.
- Lock down unintended access points.
- Reconcile updated files with precision.
- Finish it all in under an hour—while researchers kept working.
“It’s times like that, when you can take panicking researchers and get them back in business quickly, that you make us look like a hero. Starfish is worth its weight in gold.”
Starfish brings daily wins
Starfish didn’t just save the day during the migration. Todd’s team now uses it every day to:
- Search and discover data with Starfish’s querying abilities:
When a returning faculty member needed access to a decade-old library, Starfish helped them find it within minutes using it’s query feature and easy command-line interface. The actual query took less than two seconds to scan over a billion files. - Manage cold data and storage costs with Zones:
SCI archives inactive data to cost-effective disk and S3 targets instead of relying on slower tape systems. Starfish makes identifying that data painless via its advanced analytics and then enabling researchers to tag the data that needs to be archived, using Zones to limit their access to their own files. Starfish’s data cost calculating capabilities make the savings readily apparent to researchers. - Replace “find” with speed and scale:
Todd noted that Starfish is “so much faster than the old find command, especially at our scale. It turns hours into seconds.” With indexing and parallelism, it turns tedious tasks into fast, efficient processes.
Final thoughts
Migrating billions of files, managing changing UIDs, keeping researchers online, and avoiding data chaos—Todd Green and his team at SCI pulled it off with Starfish. What could have been a multi-day crisis became a one-hour fix, thanks to smart tools and a little bit of data sleuthing.
If you’re managing complex storage environments, handling large-scale migrations, or just want to avoid your own “open window” surprise, Starfish might be the hero you didn’t know you needed.
Want to learn how Starfish can help your team? Let’s talk.
Big thanks to Todd Green and the team at the University of Utah’s Scientific Computing and Imaging Institute for sharing this story—and for making IT look like magic.
