michelangelus - Fotolia

Startup Reduxio helps research company reduce data

La Jolla Institute for Allergy and Immunology gives startup Reduxio a shot at managing research data, with features in the array such as BackDating and NoDup.

Tasked with keeping hundreds of terabytes -- and soon, petabytes -- of data forever, La Jolla Institute for Allergy and Immunology IT director Michael Scarpelli was looking for the most efficient and affordable storage he could find.

LJIAI, based in La Jolla, Calif., is a nonprofit biomedical research institute that studies the immune system's response to infectious agents. It includes research from 25 laboratories, and data storage is crucial to their work.

Scarpelli said the institute has approximately 400 users, and the bulk of its storage is for "tons and tons of little files" of informatics data. Scarpelli said LJIAI has approximately 150 million files stored, taking up between 600 TB and 1 PB, as capacity nearly doubles every year.

"We retain data forever," he said. "We're not worried about legal discovery. If somebody makes a discovery 20 years down the line, we want to be able to say, 'We have part of that here.' So, our data [growth] is a continual problem."

Scarpelli said the institute looks for top-of-the-line storage, seeking the best system available that meets its budget. "We're pretty mercenary in what we grab," Scarpelli said. "Our budget is lean, so typically, high-end storage is out of our price range."

LJIAI had been a Dell Compellent customer for years, but switched to high-end JBODs consisting of ZFS storage software for a few years before installing Nimble Storage. When Scarpelli had a chance to test newcomer Reduxio Systems' hybrid flash array, with features such as BackDating and NoDup, he gave it a try.

Scarpelli said he was happy with Nimble for virtual machine storage before his VAR recommended taking a look at Reduxio when it came out of stealth in September 2015. He started off by putting an archive of an email server, scratch volume for informatics and LUNs that boot the informatics cluster on a Reduxio HX550 array system, but found it was ready for more.

"We weren't in the market for something new," Scarpelli said. "I didn't feel like we had a specific need, but what we saw looked compelling enough that we made it work. We brought in a test unit, ran it for a few weeks, tuned the network and blazed through stuff. It has strong dedupe, the IOPS we wanted and the rollback feature."

Reduxio's features aren't as exotic as they sound. BackDating is a form of continuous data protection that puts timestamps on a file's metadata, so the array can roll back to a certain point in time without taxing the system. That allows admins to clone any volume to any point in time for data recovery or application purposes. NoDup is global inline data deduplication and compression designed to extend the footprint flash storage. It reduces data across cache, volumes, clones and history.

An HX550 array holds eight 800 GB solid-state drives and up to 16 2 TB 7,200 RPM SAS drives for a maximum of 38.4 TB of raw capacity in a 2U chassis.

Through a grant, LJIAI set up the Immune Epitope Database and Analysis Resource (IEDB), which includes a website that serves as a warehouse for thousands of research reports. Scarpelli has Reduxio in mind for the database's back-end storage.

"It's customer-facing, so it's important for it to be up and running smoothly and quickly," Scarpelli said of the IEDB site. "It has a lot of VMs attached to it, and we do replication and failover to make sure it's always running. We'll probably move a lot of VMs on that to Reduxio. It seems to be cruising through what we're throwing at it so far, and we want to take better advantage. The services we have on [Reduxio] now aren't taxing it much. We're probably going to move higher-impact data to it."

Scarpelli said putting higher-impact data on the Reduxio will allow LJIAI to take better advantage of Reduxio's BackDating. "It's super easy, especially if you're loading [Microsoft] SQL Server on that," Scarpelli said. "You can roll it back to the second. That sounded nice to us. Why make restores hard, if they don't have to be? We're also getting high IOPS, and if you're hosting VM LUNs, you want the system responding quickly."

He said the HX550's clean user interface and ease of use also helps his team, which has the equivalent of 3.5 full-time people to run all IT outside of the help desk.

"When we look at products, we look at the user interface, because I don't have time to read a 500-page manual and then hunt around to find the settings I need," he said. "We lean toward something that is user-friendly."

Next Steps

Data storage startups on the radar in 2016

Deciding between hybrid and all-flash arrays

All-flash begins to outpace hybrid storage arrays

Dig Deeper on All-flash arrays