“There’s a fundamental difference with Qumulo: the ability to linearly scale capacity and performance at the same time.”

Nick Rathke, Assistant Director, Information Technology

ABOUT

The University of Utah’s Scientific Computing and Imaging (SCI) Institute is an internationally recognized facility specializing in visualization, scientific computing and imaging analysis.

USE CASE
  • Scientific Computing
  • Research
  • Life Sciences
REQUIREMENTS
  • Scalable performance as well as capacity
  • Data management through visibility
  • Simple, transparent pricing structure

Visionary Research Institute Cuts Image Processing Time with Qumulo

Storage I/O bottlenecks were dramatically slowing imaging projects for the University of Utah’s SCI Institute, while lack of insight into data usage hampered effective capacity management. Only Qumulo, an enterprise-proven file storage system could cut processing times from months to days, and deliver the analytics they needed to make smarter decisions about managing their data. For the SCI Institute, choosing Qumulo meant turning their black-box storage into a highly visible resource.

Photo of medical staff consulting an imaging report

Massive Data Files Deliver Massive I/O Bottlenecks

The University of Utah’s Scientific Computing and Imaging (SCI) Institute is an internationally recognized facility specializing in visualization, scientific computing and imaging analysis. Like any research organization that deals with scientific imaging, it’s confronted by massive data files – and equally massive processing and capacity challenges.

“Our projects are much too large to fit in memory, so I/O performance is critical to our ability to quickly process data sets,” observed Nick Rathke, Assistant Director, Information Technology for the SCI Institute.

As an example, the SCI Institute, which focuses heavily on medical research, conducted a cutting-edge analysis of rabbit retinal structure as an analogue to human retinas. Using electron microscopy, thin slices of the retina were scanned, and then assembled into a high resolution, 3D image for researchers. Unfortunately, image processing on an original 20TB data set like this typically took upwards of three months or more.

“We have massive parallel supercomputers that are starved for data because we can’t get anywhere near saturation on our disk I/O,” said Rathke. “We’ve purchased scale-out NAS systems that expand capacity, but not
performance, so we’re stuck with architectural decisions made years ago based on much smaller data sets.”

As the size of imaging data sets have grown, so too has the impact on storage capacity.

“There’s a fundamental difference with Qumulo: the ability to linearly scale capacity and performance at the same time. Every time I add a node, I’m adding network bandwidth and I/O operations, not just a tray of disks.”


Nick Rathke, Assistant Director, Information Technology

Visibility offers a path to better management

“When we run out of capacity, the direction from higher up is inevitably ‘just delete old data’,” Rathke said. “But which old data? There’s a big distinction between data that’s old and data that’s important, and I can’t tell which is which without running lengthy manual reports.”

Given this lack of visibility, there’s also no way for Rathke’s team to work directly with users on storage management. “I can’t easily tell them how much they’re using, I can’t dispute the importance of a file that hasn’t been touched in years, I can’t track allocations – it’s an extremely painful process.”

In short, the SCI Institute’s massive data files and high performance compute cluster demanded an equally high performance storage solution, delivering large, shared single pool capacity, multiple parallel processing paths and the visibility necessary to understand what was happening with all that data.

“Cheap and cheerful storage, like the kind we previously relied on, just doesn’t cut it anymore in our data-rich environment,” Rathke explained.

BENEFITS
  1. Slashes image processing times from months to days
  2. Eliminates throughput and IOPS bottlenecks
  3. Scales performance linearly along with capacity
  4. Delivers real-time visibility into data usage for intelligent archiving
  5. Allows pay as you grow, modularly expanding capacity and performance

For SCI, scale is a question of performance as much as capacity

To meet its needs, the SCI Institute installed Qumulo – a modern file storage system that is fast, flexible, scalable and has the real-time analytics necessary for visibility into data usage and performance at petabyte scale.

“Other vendors claim to ‘scale up’ storage, but there’s a fundamental difference with Qumulo: the ability to linearly scale capacity and performance at the same time,” says Rathke. “Every time I add a node, I’m adding network bandwidth and I/O operations, not just a tray of disks, and that’s critical for us.”

New nodes automatically and non-disruptively join the cluster to expand the SCI Institute’s single pool of storage – adding scale without adding downtime. And, that’s a real benefit with grant based research funding, giving Rathke’s team the ability to incrementally fund and expand storage capacity and performance as needed.

Photography of medical imaging on a computer monitor

Turning Storage from Black Box to Precious Resource

The move to Qumulo paid immediate dividends, slashing processing of image data like the rabbit retinal scans from 11-14 weeks down to only nine days. Software enhancements delivered through Qumulo’s two-week development and release cycle are shortening those processing times even further, as are the performance improvements each time a node is added to the Qumulo cluster.

This ability to scale I/O performance is a critical enhancement to the SCI Institute’s productivity, allowing the central storage cluster to finally keep pace with the performance of the HPC server cluster.

But the real eye-opener for Rathke was Qumulo’s real-time data analytics, which deliver instant visibility, enabling users to have a real conversation about storage resources. “Everyone thinks their data is critical, but now when someone disputes the importance of a project that’s due to be archived, I can quickly pull up the dashboard to show that it hasn’t been touched in years. That adds clarity into an otherwise murky storage decision.”

Conversely, the real-time visibility enables Rathke to also show that sometimes, no matter what the age of a file might be, it can still represent a data set that gets accessed regularly.

“It’s impossible to make smart decisions about data when your storage is a figurative black box,” notes Rathke. “Qumulo lets me know in an instant how the data is used, who touches it and how often – so storage is no longer a technical issue, but a management decision.”

“Other storage solutions require much more integration, much more set up, much more testing, much more everything, honestly. It was extremely different with Qumulo, for the positive.”


Nick Rathke, Assistant Director, Information Technology

A partnership built to last

For the SCI Institute, managing that storage has never been simpler. When another group considering Qumulo within the University asked about its ease of management, Rathke had a ready answer: “I told them to have senior staff figure out the architecture and configuration, then just throw the daily management over the fence to the student help desk, because from a management standpoint it’s that easy – a true ‘fire and forget’ solution.”

Throughout the implementation process, Qumulo’s dedicated customer support team worked in lock-step with Rathke’s group to ensure smooth deployment and growth. “Their flexibility and willingness to help us make the right decisions – not just up front, but going forward – has been key to keeping us on track,” he said.

In the future, the SCI Institute plans to add an additional QC24 node to its cluster every month or two, and looks forward to exploring more features and future software enhancements. In particular, Rathke’s team will be exploring Qumulo’s REST API to give users direct insight into storage usage through a customized user dashboard.

“We used to say ‘just add more capacity’, but what you’re doing these days with storage is so much more complicated and mission critical than it was 10 years ago – and the volumes are so much larger – that for us to effectively grow we need to consider storage a finite and precious resource, and manage it accordingly. That’s what Qumulo gives us.”

Want to learn more?

Give us 10 minutes of your time, and we'll show you how to rethink storage data.

GET A DEMO