itestro - Fotolia

News Stay informed about the latest enterprise technology news and product updates.

Cloudian HyperStore debuts in new scale-out arrays

SDS vendor Cloudian announced scale-out appliances for petabyte storage, along with Cloudian HyperStore software enhanced with self-service data protection.

Cloudian Inc., which bills itself as a software-defined storage vendor, today upgraded the hardware that its software runs on as it tries to move into the enterprise.

Cloudian's new modular array series scales to 384 PB in a 100-rack footprint and is built with hot-swappable micro-node servers, so the array keeps running even when a node fails.

The FL3000 series, built on the company's Forever Live platform, starts at 3.8 PB in a single rack and scales into hundreds of petabytes for the software-defined data center. The arrays operate on Cloudian HyperStore software-defined storage (SDS), which has been enhanced with new, self-service data protection policies for cloud storage.

The FL3000 appliances come with eight nodes in a 3U chassis and modularly scale as more capacity is needed. Each node holds 128 GB of memory, Intel Xeon E5-2600 processors and dual flash drives to optimize metadata and run the operating system. The data storage is provided in 60-drive, 4U capacity expansion storage shelves. One expansion JBOD shelf contains 60 8 TB Serial-Attached SCSI (SAS) hard drives for a maximum 480 TB on each shelf. The micro-node server has two Gigabit Ethernet network ports.

The FL3000 arrays can scale up to 384 PB within 100 racks. Cloudian will continue to sell its HSA 1000 and HSA 2000 appliances to small and medium-sized customers, while the new FL3000 targets enterprise-level companies.

"This is a total overhaul," said Paul Turner, head of product management and business development at Cloudian, based in San Mateo, Calif. "This is for the petabyte-scale environment. It's very dense and we have separated the data drives from the compute, so that you can pull out compute nodes and replace them on the fly. No single component can bring the system down. Every failure is isolated."

This is a total overhaul … No single component can bring the system down. Every failure is isolated.
Paul Turnerhead of product management and business development at Cloudian

Turner said replacing nodes does not require any data movement. Cloudian's software uses erasure coding or replication to protect data.

"One of the biggest challenges is that once a node fails, you have to move a lot of data around," he said. "In this design, you don't have to. We rebalance live data across all the other drives in the system. Our software spreads data across the drives and nodes."

Chris Wilder, analyst for cloud services and enterprise software at Moor Insights and Strategy, said Cloudian's enterprise arrays now bring it into competition with the large storage vendors. Not only is the enterprise storage hardware market crowded, it is also an area where there has not been much growth.

"The modular approach they are taking is pretty cool," Wilder said. "It's hot-pluggable and removable components. But I see challenges with the appliance as a service model. It's an uphill battle. The hardware business is a hard business to be in. You start competing with the big guys."

Cloudian also rolled out HyperStore software. The new self-service storage capabilities for Cloudian HyperStore 5.2 allow administrators to set data durability and protection policies per application. The IT department still manages data distribution and protection at the data center level.

"Our belief is you have to move from administrator-control to more self-service, like the cloud," Turner said.

Next Steps

Guide to working with software-defined storage

Answering common software-defined storage questions

Defining SDS is still a tough task

Dig Deeper on Cloud object storage

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

Does Cloudian's model of pre-packaged, software-defined storage appeal to you more than running it on your own hardware?
Well, it was good enough in the past to say that Cloudian's HyperStore software ran on x86 industry standard, commodity storage servers of your choosing, but times may be changing with this announcement by Cloudian. To say your storage is "software defined" is one thing, but supporting it on virtually any x86 storage server can create support and troubleshooting headaches for Cloudian, and its customers. By "owning" the storage server hardware with their line of HyperStore Appliances, Cloudian is betting that SMB and Enterprise customers would rather "rack'em and stack'em" than "roll their own" storage clusters. That said, a DIY HyperStore software install on your own storage servers is still a workable solution for those who are so inclined, but the ease of installation and reduced configuration time of the HyperStore Appliances makes a lot of sense. I agree with Chris Wilder that Cloudian's move into the storage server hardware market may pose some challenges, but Cloudian has not developed proprietary firmware and/or controllers for its HyperStore Appliances. Rather, it has chosen to work with several well-know OEM storage server manufacturers who build to Cloudian's specifications. This is a lot different that buying storage server products with a lot proprietary code and firmware that are expensive to acquire and own. Cloudian's HyperStore software is supported on the CentOS Linux operating system and incorporates Cassandra, Redis and Puppet along with Cloudian's HyperStore data storage engine. Cloudian is also certified to run the Hortonworks Data Platform directly on Cloudian cluster nodes, which eliminates the need for a separate Hadoop storage cluster. Cloudian is smartly sizing up the broad market for object-based storage that is fully AWS S3-compatible with its line of HyperStore Appliances.
Cloudguy, one of Cloudian's key offerings is that it's certified to run the Hortonworks Data Platform directly on a Cloudian cluster so you don't need a separate Hadoop storage cluster. Thank you for pointing that out:). It's a key benefit for companies that are highly focus on data analytics.