Red Hat Inc. is seeing a quicker pace of adoption in the initial three years for its open source-based storage...
than it did for enterprise Linux, according to Ranga Rangachari, vice president and general manager of the company's storage and big data business unit.
Red Hat recently rebranded its storage offerings. Red Hat Gluster Storage is based on the company's October 2011 acquisition of Gluster Inc., which sold and supported a commercial version of the open source GlusterFS file system. Red Hat Ceph Storage is rooted in the company's May 2014 acquisition of Inktank Inc., which based its commercial product on open source Ceph distributed block, object and file storage.
Open source communities continue to develop and enhance GlusterFS and Ceph storage software. Red Hat updates its products in sync with new versions of the open source software, typically on an annual basis, and provides commercial support.
In a recent interview with TechTarget, Rangachari discussed the market for open source and software-defined storage, the best fits for Red Hat Gluster versus Ceph, and new capabilities in the works for Red Hat's storage products.
How would you characterize the appetite for software-defined storage during the past year?
Ranga Rangachari: Last year, it was more a question of evangelizing the concept. What is software-defined storage, and where does it come into play? Now that's behind us. Now we have customer conversations like: 'The only way I can create a storage environment in this new emerging infrastructure is through a software-defined storage solution. Red Hat, tell me how you can help me with the workload that I'm running.'
The nature of conversation has dramatically changed over the last 12 months, and we see that in terms of adoption over the last few quarters. The storage portfolio is being driven as part of some of the large deals that we do.
I feel very good about the trajectory, not just where Red Hat is, but also where software-defined storage is headed overall. This is also backed up by a lot of our conversations with analysts like IDC, Gartner and Forrester. You're now seeing this go from what I would call the lunatic fringe of early adopters to the mainstream, where people are actually starting to build infrastructure for the long haul for storage.
What do you tell customers about the main value of software-defined storage?
Rangachari: It's decoupled. The traditional way is a proprietary piece of hardware with the software in it. If the hardware and software are tightly integrated where you just cannot take it apart, they are in a way locked in, not just in terms of cost and everything else but also in terms of innovation. They cannot upgrade their hardware because it's in a four-year depreciation cycle. With software-defined storage, you run the software on top of industry-standard hardware. Today it could be Dell, Cisco. Tomorrow it could be HP, Supermicro. It gives them complete flexibility to run the same piece of software on varying hardware.
With traditional vendors talking more about running software on commodity hardware, how do you see your competition stacking up today as opposed to a year ago?
Rangachari: There are existing companies that were in the space of proprietary solutions, and now all they're trying to do is basically retrofit or try to come over in a kluge way and just rebrand it as software storage, when it wasn't designed from the ground up to be software-defined storage.
Besides Red Hat, I look at quite a few companies, startups primarily, who are doing quite a bit of innovative work. They're designing solutions, much like Red Hat, that are built from the ground up for a software-defined infrastructure, not taking an existing piece of a product and then pulling their software out, because that will not withstand the test of scale and reliability.
What do you think of traditional storage vendors such as EMC and IBM that are talking about decoupling their software and hardware?
Rangachari: My view of that is, easier said than done. The intentions are great. But I think the jury is still out on whether they've achieved a modicum of success in their implementation and customers saying, 'Yup, I like it. Keep it cheap.' It's got to transcend marketecture. It's got to have real customers using the solutions in a pure software-defined storage fashion.
Is there a killer app that you think will drive users to open source software-defined storage?
Rangachari: One of them is OpenStack. There is a huge need for storage in OpenStack environments, and we are getting called to those dances.
When do you predict Red Hat will achieve the success of Linux on the storage side?
Rangachari: I wish I had a crystal ball on that one. That's a tough question. I think in the next couple of years or so, we'll be there, just given the trajectory we're on. I can't back it up with any data. This is just purely based on my conversations with customers, partners, service providers and such, that this is no longer something that they'll look at when they have time. Right now, they're saying, 'I need to move onto this pretty quickly.'
What are the big new features coming with your Gluster and Ceph products?
Rangachari: There are a lot of things in the works. We are going to be doing quite a bit around tighter affinity with OpenStack, not just where we are today, but with things like Manila [shared file system service]. We've been involved in that along with NetApp and IBM and others for almost a year and a half now, and you're going to see the fruits of that labor pretty soon. The other area that we are paying very close attention to is a project called Trove, which is database as a service. The four basics today around OpenStack are file, object, block and database as a service. I think we've done a very good job on the block and the object. But we need to do quite a bit of work on the file and the database as a service. Our goal here over the next few months and quarters is to start to make progress on all those four fronts.
The other area that you'll see us eventually make some noise around is the concept of hyper-convergence, where you can have compute and storage run on the same box. The good news for us is that with our Red Hat Enterprise Virtualization platform and with our storage platforms, we can now have a truly integrated product that helps customers with their hyper-converged story. Where we are getting a lot of positive feedback is ROBO, which is remote office branch office, and specifically around the finance and the retail vertical.
Red Hat supports Ceph's block and object storage. Will Red Hat support Ceph's file storage?
Rangachari: Yes, when it's enterprise-ready, when we think that we can put the Red Hat badge behind it. But, it's in early stages in the [open source] community.
Will Ceph FS eventually usurp the Gluster file system?
Rangachari: Every other company has different products for different workloads. If you take a view of the workload aspect of this, it's not going to be at the expense of one or the other. I think that there is enough of a workload use case out there that each of these solutions can be ideal for certain workloads.
When you talk to customers, how do you position Gluster versus Ceph?
Rangachari: All the conversations that we are seeing with our customers today are workload-driven. A typical workload [example is], 'I have a massive project internally around OpenStack. I need to provide storage services within OpenStack, but primarily around block, because that's where all the virtual machine [VM] storage resides.' In that specific example, the Ceph storage product is an ideal fit.
Contrast that with when a customer says, 'I've got a huge analytics project going on,' whether it's Hadoop or Splunk or any of the log files, where there's a tremendous amount of unstructured data that's created. They look to us and say, 'What do you think is the best fit for this type of a workload?' In that case, it's the Red Hat Gluster product line today.
Analytics, enterprise sync-and-share, and enterprise virtualization are ideal fits for the Gluster product. At the other side of the spectrum, you have Ceph, which is an ideal fit for cloud infrastructure. This is classic VM storage with OpenStack and object storage. If you look at it from a file, object and block standpoint, object storage is expected to be the fastest-growing part of the market.
Rich media is another area. If it's file-based media, it's predominantly a Gluster-based solution. If it's object-based archival and rich media, it's primarily a Ceph-based solution. There is a very clear line of demarcation today.
Is Red Hat Gluster adequate for all workloads suited to file storage? What are the file system's weaknesses at this point?
Rangachari: One area where we typically tell our customers there are better solutions in the market is if somebody were to deploy databases, Oracle databases, those type of things, where it's very much scale-up. It's not scale-out unstructured data, where things like NoSQL come into play. That's where we shine the best.
What are the main gaps in Ceph right now?
Rangachari: Customers obviously want a better user interface. The Calamari project is a great start, but that's one area where we will continue to invest in. [Another] area that customers have asked about is reference architectures. A customer can say, 'I'm thinking of deploying in a Dell server environment. Give me some ideas of what I can expect in terms of performance, what are the parameters that I can fine-tune.' What we are doing right now is continuing to publish a wide range of reference architectures for all the leading server platforms, so we can take as much guesswork out of the system as possible.
Where does Red Hat stand with hardware partnerships?
Rangachari: We announced a relationship with Dell [for Ceph]. It's kind of the next wave of the relationship. We already had an existing relationship with Dell on OpenStack. On OpenStack, we've also announced a relationship with Cisco [for Ceph]. With Fujitsu in Europe, [there's] a hardware platform that actually has Ceph built into it. The other more recent announcement is SanDisk. Even before our acquisition of Inktank, they were working with Inktank for over a year to embed Ceph as part of the InfiniFlash appliance.
Red Hat strives to copy success achieved with enterprise Linux
Red Hat Storage's early users report promising results
OpenStack Manila file service progresses
Learn more about system optimization and containers