This content is part of the Essential Guide: Red Hat Storage Server seeks to mimic success of enterprise Linux
Get started Bring yourself up to speed with our introductory content.

Software-based storage is 'way of the future,' Wikibon CTO predicts

Wikibon CTO says software-based storage using commodity hardware, drives and controllers is the way of the future; Red Hat will help drive down costs.

Listen to the entire podcast series about Red Hat’s chances with its storage server software

Red Hat's open source software-based storage holds appeal mainly for passive data

GlusterFS storage software will turn the market 'on its head'

Red Hat Inc. was one of the major forces in driving acceptance of Linux in enterprise IT shops, but the company will need help from other major players to achieve success with its new push into software-based storage, according to one leading industry analyst.

David Floyer, co-founder and chief technology officer at Wikibon, a Marlborough, Mass.-based research and analysis firm, said Red Hat will play a role in driving down the cost of storage features such as replication and thin provisioning as the IT world moves to what he calls "software-led storage" based on commodity hardware, disks, flash drives and controllers.

Floyer called software-led storage (more commonly known as software-defined storage) "the way of the future" and said he expects other major industry players to also play a significant part in bringing about its acceptance.

In this podcast interview with TechTarget senior writer Carol Sliwa, Floyer shared his views on Red Hat's chances at commercializing the open source GlusterFS distributed file system and the OpenStack cloud computing and storage platform, as it did with enterprise Linux.

Can Red Hat bring about mainstream acceptance of software-based storage, in other words, storage software that doesn't require any special hardware?

David Floyer: I'm not sure it's Red Hat that's going to bring about that acceptance. I think software-led storage is the way of the future. It's going to be based on commodity hardware, commodity disks, commodity flash and commodity controllers, mainly Intel controllers, and it's going to be the software that's the differentiation of that. So, I think the industry in general is going to drive very strongly toward that model. Red Hat being part of the open source [community] is one of the drivers toward that model. Clearly, at the moment, there is enormous value from the additional software added on by vendors in their storage arrays. Replication, thin provisioning, etc., are all individual features that cost money. Over time, the cost of those features will go down very significantly, and Red Hat will be a factor in bringing those costs down. But it's not just Red Hat.

Do you think Red Hat can commercialize GlusterFS in the same way it did with enterprise Linux?

Floyer: GlusterFS is a great product. It's clearly a good clustered file system. It handles thousands of clients and petabytes of data and offers a lot of storage capabilities, such as building blocks over InfiniBand, for example, or RDMA [Remote Direct Memory Access] or TCP/IP interconnect, and has a lot of the fundamental functions. I also believe it has some challenges as well in becoming a standard. It needs to be able to allow itself to be managed by the application as part of the compute, the storage, the networking resources. For that, it needs to be managed through an API. And the level of management has to be quite detailed and sophisticated. It's got to be able to allow bandwidth to be managed, [and] response time, IOPS [and] latency to be managed. And it needs to manage minimum and maximum and burst storage levels -- resources from an application perspective. Now that's a way to go from the way that cluster file system is at the moment. You can't have scripting if you're going to automate in this sort of way. It has to be through APIs. Obviously, cluster has some multi-tenancy capabilities. It also needs things like [a] virtual cluster. It needs a fair degree of work to bring it up to be a full cluster system of the future.

To which vendors and which types of storage will the Red Hat storage model pose the greatest threat, assuming the software matures?

Floyer: Oh I think it will mature, and I think clearly its area of initial acceptance will be in the file-based systems, maybe in some of the archiving systems. And that seems a slightly peculiar thing to say in the sense that it's obviously a system that's used in some of the HPC [high-performance computing] systems at the moment. But those are the systems where its ability to cluster things effectively at a low cost will mean more than it will in some of the other areas. So, I would predict it will be first used in the simple file-based systems and allow particularly cloud storage providers to provide those sort of services economically.

Do you think Red Hat can do for OpenStack, and especially OpenStack Storage, what it did for Linux?

Floyer: I believe Red Hat is in a good position to make OpenStack successful. I think they have had a very good start. I was at the last OpenStack conference, and the degree of directness working from real requirements of the users, and the good relationship between users and vendors and their ability to support OpenStack, I thought was very good indeed and boded very well for the future. They're not the only players here, of course. HP [Hewlett-Packard] is strongly backing it, and also IBM. So, there are some major players behind OpenStack, and I believe that, same as in Linux, it needs more than just Red Hat to make it successful. And I think they have a good shot at being one of the dominant platforms for high-volume computing in the future.

When it comes to the storage, it'll be interesting to see if they can get Gluster accepted as part of that. But I think that with Swift and Cinder, those are solid storage foundations. A lot of work is going into Cinder in particular to do with, for example, the ability to provide some of the manageability capabilities, bandwidth, response time, IOPS and latency that I mentioned before and some of the ability to be managed by an API. There needs to be a level of management that is orchestrating this and automating this at the application level, and storage then is just a part of the system, and the real management comes from this API level for the application. That has to be the mindset. And one of the biggest challenges that Red Hat and the others have may be providing that high-level orchestration.

To achieve success with OpenStack, what will Red Hat need to do moving forward?

Floyer: The key to success with OpenStack and with OpenStack Storage is to have the ability to orchestrate and automate at a control level, which is in line with the application itself. The days where you guess what the application is doing or infer from it need to be over. It's got to be able to provide the IOPS, networking resources [and] the compute resources required to guarantee the performance of that application, and be able to adjust those resources up and down in real time with the right level of security and compliance. That to me is the single biggest challenge of that OpenStack environment, and the vendors and Red Hat who are contributing to it.

Can they get to that level? It's all very well to do that for a single application. If you have just one application running on Facebook or whatever it is, then it's easy to do because you can put in an orchestration that is specific to that application. But to make it general purpose and make it a really effective way that enterprises can run the many applications they need to run in that environment, then it has to be pretty flexible and pretty effective. It's significantly more complex than just an OS, and it's significantly more complex than an OS for a single big application. If they can do that, they will really have the opportunity to make OpenStack the default standard for large-scale computing in the next decade.

Dig Deeper on Software-defined storage

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.