Storage software increases options for developers
August 14, 2012 —
(Page 1 of 2)
Related Search Term(s): Ceph, EMC, NetApp, storage
Today, storage is about big boxes, from companies such as EMC and NetApp. But a number of new companies have taken up the banner of storage systems as software. With their help, developers can free themselves from the ties that bind them to on-site corporate storage systems.
Bill Roth, vice president of marketing at Nexenta and formerly at Sun Microsystems, said that new storage options from his company bring some old Sun technology back to the forefront. Sun's Zetabyte File System was pushed into the open-source community even before Oracle had taken over Sun. But when Oracle acquired the company, ZFS was somewhat shunted to the side; Oracle also has the BTRFS project, which seeks to create a similarly massive file system for scaling storage over the next decades.
While Oracle still offers ZFS in its Solaris products, Nexenta uses a fork of OpenSolaris known as Illumos to build a storage-focused Solaris distribution based on ZFS.
“In many ways, NetApp and EMC are saddled with the innovator’s dilemma: They've got old file systems on proprietary hardware,” said Roth. “We're seeing success with open-source software, with new file systems and commodity hardware. I don't think the legacy vendors are going to be able to keep up.”
Another upstart software-based file storage system is Ceph, which grew up in the academic world. Designed by Sage A. Weil while he was attaining his doctorate in computer science at the University of California, Santa Cruz, Ceph is a flexible storage solution that can host objects, file systems or blocks. Based on the CRUSH algorithm, it is able to accommodate multiple types of storage systems on the same platform, making it stand out from the rest of its ilk in the OpenStack world. Originally designed as a scalable storage system, Ceph was recently submitted to OpenStack as an alternative for OpenStore, the OpenStack Object Storage system.
Today, Ceph includes a number of sub-projects that enable the system to scale to the petabyte range. In June, Ceph received corporate backing in the form of Inktank, a company formed in order provide to service and support of Ceph.
Ross Turk, vice president of community at Inktank, said that Ceph is unique because of its flexibility and robustness. “The interesting thing about Ceph is that it's the only storage system I've seen that has been designed from the very first step to have no single point of failure, and the only one designed for failure as the norm,” he said. “For every 1 million nodes, a bunch will fail every day. The way the CRUSH algorithm works gets around a lot of the problems that are limiting the scalability of stuff that starts with the file system and builds on top of it.”
For developers, these new storage systems increase the flexibility of their environments, and give them more deployment options. Currently, large storage systems from EMC or NetApp live in corporate data centers. They can't move because they are tied to extremely expensive hardware. Thus, developers can use these systems internally, but cloud-hosted apps are either unable to use such storage, or must do so through a high-latency connection from the cloud into the data center.
With Ceph, Nexenta and other software-based storage systems, that on-site data store can be replicated in a cloud, using the same storage software that exists on premises.
Bryan Bogensberger, president and COO of Inktank, said, “Ceph is changing how people are able to deploy IT infrastructure because of the fact [that] it’s software and it’s architected the way it is.”
Such software-based storage options also give developers the opportunity to rewrite storage interfaces the way they see fit, said Nexenta's Roth. “We have open APIs. We do have our own user interface, or you can write your own if you don't like ours,” he said.
“If they want to build their own cloud, they can throw up OpenStack and use this as a storage base underneath, or they'll have their storage structure effectively virtualized. We've seen a lot of people putting Hadoop on top of Nexenta, because even though the data's replicated, failure can still be very costly. How do you limit that? You put ZFS under it to make it even more powerful.”