Ivelin Radkov - Fotolia
The change in focus of Red Hat storage from stand-alone Ceph and Gluster products to OpenShift Container Storage, hybrid cloud object and hyper-converged technologies raises questions among customers.
Multiprotocol software Ceph will soon replace file-centric Gluster at the foundation of OpenShift Container Storage (OCS). The switch gives Red Hat storage customers more scalable S3-based object storage, along with file and block options. Gluster will still play a small role in the Red Hat storage portfolio as the virtualization option of the Red Hat Hyperconverged Infrastructure product.
We caught up with Ranga Rangachari, vice president and general manager of Red Hat storage, to discuss the rationale behind the decisions, the potential repercussions for customers and the future of Gluster storage.
Why did you shift the Red Hat storage strategy from focusing on stand-alone Gluster and Ceph storage products to OpenShift Container Storage, hybrid cloud object storage and hyper-convergence?
Ranga Rangachari: It's a combination of a few things. The first is the market shift. A little over six years ago when we entered this space, things like hybrid cloud and even something as basic as software-defined storage weren't yet mainstream in users' minds. People were still equating storage with a piece of iron that gets walked into your data center.
With the acceptance of a hybrid cloud strategy across all these different organizations, the need for software-defined storage becomes more and more pronounced. You cannot take a piece of hardware and move it into a public cloud. It's just physically impossible. So, what that did was elevate the conversation to software-defined storage.
The other aspect is our initial entry into the market, even with Gluster in the early days, was around cost. Red Hat 1.0 was built around, 'Let's go attack a market and use commoditization and cost advantages as our value proposition.' That's changed dramatically, where it's now on innovation.
When you shine a spotlight on how we can help our customers innovate within the context of the hybrid cloud story, the center of gravity shifts away from stand-alone storage -- where it's, 'I can do it for 3 cents less than the competitor' -- to one of how can it help you excel in the hybrid cloud journey, where innovation is what's driving the conversation.
This is not that we woke up on Jan. 1 and said, 'Hey, we're going to flip the switch.' We followed this trend over the last few years. And we felt that now's the time, as the hybrid cloud option becomes more and more mainstream, to fine-tune our product portfolio to have the major focus on hybrid cloud storage and the minor focus on the stand-alone play. It's almost an inverse of what you heard from us over the last three or even two years.
With the change from Gluster to Ceph in the OpenShift Container Storage product, what do you say to customers who were already using Gluster with OCS?
Rangachari: Let's say I'm a developer. I'm building an app on OpenShift. The underlying storage substrate is completely transparent. They don't care, and they don't need to care, because we will absolutely provide the exact same level of functionality within the OpenShift 4 framework as we did within the OpenShift 3 framework. This is an integrated product in all senses of the word. It's not that somebody is going to be running OpenShift 4 storage separately somewhere in the ether. It's all going to be tightly integrated into the OpenShift Container Platform.
For the on-prem folks who are running OpenShift 3 today, there is a migration toolkit that we are providing our customers. They'll start to migrate their infrastructure from an OCP 3-based cluster to an OCP 4-based cluster.
Will there be a cost for customers in terms of time and additional storage capacity for the migration?
Rangachari: There should be no additional expense for somebody to move from OCS 3 to OCS 4 in terms of procurement cost, subscription cost or whatever else it might be. So, if I'm running OpenShift Container Storage 3 in a certain configuration and move to OpenShift Container 4 in a similar configuration, there should be no reason why the customer has to pay extra money.
We don't charge for a release. If they're an existing customer with OCS 3, and OCS 4 comes along, that's part of the subscription.
If they choose to say, 'I'm done with my OCS 3 chapter,' they can reuse the storage for OCP 4. It's not that they've got to go buy 200 PB of new storage to make the migration. We don't anticipate any incremental tax -- whether it's extra hardware procurement, storage procurement or extra subscriptions -- because of this migration.
Is Gluster going to be phased out over time?
Rangachari: That's a tough question, because I don't know where things are going to be 15, 20 years from now. If you ask me today, is there a formal plan to deprecate Gluster? The answer is no.
There is a Red Hat Gluster Storage product in the market today. We are working on a new release for later this year. It's not that it's getting marginalized or diminished. But it is going to be focused on specific stand-alone use cases where we think it can really shine.
What are the main use cases you envision for Gluster moving forward?
Rangachari: Where we are seeing a lot of use for Gluster at a high level is classic NAS -- 'I want to provide a file share across a bunch of users' -- and predominantly a little bit of Windows and a lot of Linux environments where it's CIFS and NFS.
Ranga RangachariVice president and general manager of storage, Red Hat
In some financial institutions, we have people who are actually using Gluster for capturing big data. It's all stored in files, and they analyze it later. There are people who want to provide video streaming for their internal organizations. Those things continue to go well. So, those use cases are still appealing, and we still cater to that.
It's not that we are putting the brakes on one and pressing the accelerator on the other. The technology rationalization that we made was predicated on three key things. One is the community innovation with Rook. It's become almost a center of gravity of some of the innovation that's going on in the Kubernetes storage world. The second is the maturity of CephFS for the OCS use case. And the third is the S3 object stuff. And last, but not least, the NooBaa acquisition that we did last November is an integral part of that [OpenShift Container Storage] offering.
With OpenShift object storage and OpenStack geared toward massive scalability, are you trying to cater to the largest deployments and leaving the SMBs behind?
Rangachari: No, we are not. Let's take OpenStack out of the equation, because OpenStack tends to be large enterprises building clouds and stuff. But there is a huge correlation between the OpenShift deployment patterns and the OpenShift storage deployment patterns. OpenShift caters to a wide swath of audiences -- large customers and some small customers.
If an organization, regardless of its size, is on the container journey, and they want to build a platform that they can develop apps on that gives them the freedom to run it on-prem or on one of these three public clouds, they choose OpenShift.
Is Ceph well-suited to any size deployment, or is it geared for large deployments?
Rangachari: If you look at Ceph today, it's primarily associated with OpenStack and those types of modular implementations. But the few things that we've done with our integration with Rook, containerizing Ceph, helps us take all the heavy lifting that one could have done within the context of OpenStack.
At the core, the technology is the same. But how it's deployed, instrumented and provisioned is dramatically different. A classic example is the operator for Ceph. The first thing that it does is provision these clusters, and it doesn't need any human intervention to go in and deploy all these things. That's what the operator is all about. If you were to do an upgrade, the operator does the upgrade for you across all the Ceph clusters.
Why doesn't Red Hat support both Gluster and Ceph in OpenShift Container Storage moving forward?
Rangachari: This is where the open source community and other things come into play. If this was a proprietary software company that has 85 different storage products, [we could say] let's go do this. But let me give you an example of what we have to keep in mind. It's not just saying, 'Let's support Gluster.' It's about, 'What is the interaction between Gluster and Rook?' How do we make sure the community stays involved in that?
Ultimately, the other part of the conversation is sometimes choice leads to confusion. The customer is going to use Gluster for this, but later on want to move Ceph there. All we are doing is kicking the can down the road, to a certain extent, in offering these choices. We felt that now, where there's a migration between OCP 3 and OCP 4, this was the right time to do it.
Do you see big storage players such as Dell EMC and NetApp as your competitors?
Rangachari: For file servers. Say, if you use one of the vendor's file service products, there were customers who moved from theirs to ours, because cost was the main driver. We've seen with cloud and other things, it's an important part, but the trump card is innovation more than cost.
You originally had visions of making Red Hat storage as popular as Linux. Is that still your vision?
Rangachari: Our vision is through our storage portfolio to help our customers be successful in their hybrid cloud journey. It's plain and simple.