Red Hat Inc. expanded its move into storage last week with the integration of its storage software and its distribution of RedHat OpenStack cloud technology, as well as an updated roadmap to fill in key gaps with its year-old storage server product.
Open source gains credibility with users
Best known for its enterprise Linux distribution, the Raleigh, N.C.-based company has become a major contributor to the OpenStack community, which develops the open source software designed for enterprise use with private, public and hybrid clouds.
The integration of Red Hat Storage and Red Hat OpenStack -- which includes Swift object storage and Cinder block storage -- followed an April announcement that the company's supported OpenStack distribution had advanced from preview state to early adopter program.
"They want to do to OpenStack what they did to Linux, and I applaud them for going there because OpenStack today is more chaotic than Linux was," said Ashish Nadkarni, a research director in the storage systems practice at Framingham, Mass.-based International Data Corp. "They have a discipline internally to make the OpenStack movement like Linux, where it stays open source, but at the same time has a commercial support model."
The OpenStack announcement at last week's Red Hat Summit in Boston appeared to raise the visibility and credibility of the open source technology with some attendees. Several customers said they plan to evaluate Red Hat OpenStack technology in the coming months.
"I'm going to take a look at it; obviously it is helpful if there is a big partner supporting it," said Mohit Anchlia, a software architect at Intuit Inc., an early adopter of Red Hat Storage. He said he hasn't looked at OpenStack much because he sensed "there's still a lot of work that has to be done."
Duc Doan, director of system engineering for ETrade Financial Corp., said the company is a heavy user of open source technology and already deploys OpenStack compute software. Doan said ETrade relies on traditional storage area network storage, but will probably check out OpenStack storage options within the next few months.
"Storage still is probably the hardest thing for us to move into the cloud because of the scalability, security and migration costs associated with unstructured data. We're still looking to see how we can do it correctly," Doan said. "It's fairly easy for us to move into the cloud from the system side, but the data side is a completely different story."
Sayandeb Saha, manager of product management for Red Hat Storage, said Red Hat will be a storage provider for its OpenStack distribution and the company's storage server software will be a supported back end for the OpenStack cloud platform.
"When you are doing Swift, you are using the OpenStack APIs. The back end is Red Hat Storage," Saha said. "When you are doing Cinder, the block storage layer, we can back that up. When you're doing Glance, which is an image service, we can back that up. That is one of our use cases.
"Do you really want one storage island for files, one for object and one for block?" he asked. "If you have Red Hat Storage, we can handle object and file. That's the value prop."
Red Hat Storage currently supports file, object and virtual block and file data, but not iSCSI block-based storage. Saha said the company is working on block. Although OpenStack doesn't yet have a file service, Red Hat plans to add one based on NetApp Inc.'s file-share service proposal to the OpenStack Foundation Technical Committee, he added.
"Applications that were written to use file-based storage should be able to run in OpenStack without major modifications," Saha said. "OpenStack today is all about block. Block is great, but people have been writing file-based storage apps forever. People are trying to move away from block to some extent, so that's a natural fit for us."
Red Hat's product road map
Red Hat's endorsement of OpenStack is more than a mere extension of its support for open source software. The integration work also falls in line with the company's vision of software-defined, horizontally scalable storage that customers can deploy on the same commodity server hardware (often with JBOD devices) where they also run compute services.
Announced at last year's annual summit event, Red Hat Storage consists of: Red Hat Enterprise Linux (RHEL); the single-node XFS file system (for each Linux box); the distributed, scale-out GlusterFS file system that it acquired two years ago (which runs over the local Linux file system and pools the storage); and a console management station. Users supply their own physical servers or virtual servers.
Red Hat Storage is suited to unstructured data such as documents, images, audio and video files, virtual machine (VM) images, log files, disaster recovery copies and objects, but it is not intended for use with relational databases, according to Saha.
"The design of the technology is not optimized to handle highly transactional workloads that do a lot of small I/O and random I/O," Saha said. "Currently, it is not in our near- to medium-term roadmap to handle those kinds of workloads. While they are interesting, there are good solutions in the market for that today."
Technologies that Saha highlighted as part of the Red Hat Storage Server roadmap include:
- "Big Bend" release (version 2.1) with the theme of "high-performance geo-replication" due in the third quarter of this year. It includes updated RHEL 6.4 and GlusterFS 3.4 support; faster parallel, distributed asynchronous geo-replication; upgraded OpenStack object store support for the latest Grizzly edition; and Red Hat Network Satellite support.
- "Corbett" release with the theme of "manageability" due in the fourth quarter of this year. The Corbett release features Red Hat Storage Console support, Server Message Block (SMB) 2.0 support and performance improvements, NFS access control list support and possibly remote direct memory access over InfiniBand.
- "Denali" release with the theme of "volume snapshots" due in the first or second quarter of next year. This includes RHEL 6.5 support, RHEL's dm-thinp-based snapshots, integration with commercial backup toolsets and improved performance for small-file workloads.
Early adopters at last week's Red Hat Summit cited benefits such as cost savings over traditional storage, ease of scaling, high availability and active-active replication. David Yaffe, a systems specialist at Saskatchewan Telecommunications (SaskTel), said the price tag to store log files using traditional storage would have exceeded the budget for the entire project.
"We had to start looking at other alternatives," he said. "The reason why I started looking at Red Hat Storage a little bit more closely was that everything about it was a known technology to SaskTel. The fact that Red Hat had taken Gluster and brought it into the Red Hat family made it a very easy sell corporately, and then during our technical trial, it just worked on the first try."
Will users find a place for Red Hat OpenStack?
With the Red Hat Storage merely a year old, some customers are still trying to figure out where the software might fit into their environments and if they will find a place for Red Hat OpenStack cloud storage technology.
"Cost-wise, I think it will be a great proposition if it gives the kind of performance we need," said Pankaj Gautam, a systems architect at Macys.com in San Francisco, noting that the company currently uses IBM and HP storage arrays. "That's what we have to see; whether this is archival storage or the kind of online storage we are used to. We are used to a rich management console, and I don't see that there. So, is it the right time to go in, or do we have to wait?"
Eric Robertson, a disaster recovery architect at Northrop Grumman Corp., said he can foresee Red Hat Storage being useful for customers with small-scale applications in the low terabyte range to avoid SAN redesigns.
"It appeals to me because of the horizontal scalability and the idea of using my existing hardware infrastructure without having to buy special-purpose hardware," Robertson said.
Still, the concept of running compute and storage from the same box has some users skeptical. McMaster University in Hamilton, Ontario implemented the virtual edition of Red Hat Storage Server with its Dell Compellent disk arrays rather than commodity servers. The system is set to go live in production in December.
"The on-premises virtual version didn't come with disk, so we were going to have to provide it," said Wayde Nie, lead architect for university technology services at McMaster. "If we had to provide it anyway, then it made sense to keep it on the storage platform that we had invested in already. They have a high level of redundancy built into them."
Nie said the staff re-exposed the SAN technology through Red Hat Storage to provide the desired redundancy and active-active replication for the school's PeopleSoft Financials. The licensing for active-active replication would have been more expensive and the high availability wasn't "quite there" yet with the school's storage arrays, he claimed.
"We were looking for the features, not so much a way to get rid of our traditional storage, which we've already consolidated onto Compellent," Nie said. "The nice thing about Compellent is that it automatically tiers blocks of storage down to lower-end storage if they haven't been accessed in a while."
Nie added, "I still like the idea of consolidated storage."