Software-Defined Storage Takes Shape

It’s still hard to get a definition of software-defined storage (SDS) that covers all the ways the term is being used, and abused, in the IT industry. To some extent, the term is a smokescreen to hide the confusion in many traditional storage vendors’ product planning. The rapid commoditization of storage is threatening growth for these companies, and fueling a major loss of brand loyalty.

The cloud is largely to blame. Service providers have tested many of the sacred beliefs of the industry, and found them to be myths. Commodity drives work fine as bulk storage, at least unless you are Seagate. Enterprise hard drives aren’t measurably faster than commodity drives when measured against the performance of SSD. The result is that the era of arrays of many fast drives as primary storage is over, and flash and SSD appliances have replaced them on the leader board. RAID is dead, too!

So is software the path to happiness for these embattled vendors, and is software-defined storage thing the answer? We have to speculate on what SDS really could be to get the answer. That’s not to say there aren’t pieces of the jigsaw puzzle in the market already, but the picture is far from complete, and a bit jumbled.

The idea of SDS is actually fairly simple: Take the expensive and proprietary compute and control out of data storage appliances and virtualize it onto server instances. This separation of data services (the storing of the data) and the other functions in the storage family tree allows independent scaling of features.

In theory, the hardware in SDN is very low-level, akin to the data store device in recent object storage models. Encryption, compression, deduplication, snapshots and file systems all reside in the server farm.  Optimistically, this will create new ways to build storage solutions and open up the door for creative startups to bring their code value to the table.

Claims of existing software on existing hardware as SDS are a bit tongue in cheek. Applying the test of control plane code being in the server farm soon undercuts many such claims. Still, there are already some credible software-defined storage products.

First, anything built around Ceph likely meets the criteria. Ceph is a hot solution for unified storage, and it allows disassembly of the various control elements so that they can live on servers. The key is that data is stored on object stores, which in the end can be Ethernet-connected drives or pods of drives.

Key/data storage leads us to Seagate’s Kinetic. These are Ethernet drives with a key/data access mechanism. These drives can be built into appliances which only have Ethernet switches, or they can be directly connected to the LAN. The drives themselves are just object stores, and the control heavy lifting is elsewhere. We’ll see both  drives and the boxes to hold them in Q1 2015.

Western Digital/HGST has also tackled the Ethernet market, though with an approach that allows Linux apps to run on the drive. This can run the object store piece of Ceph, so it's a real competitor to Kinetic, but it’s clear that HGST envisions a richer software suite on its drives. Both products are so new that there’s no way to figure market acceptance and it’s possible that Ethernet drives won’t catch on in the mainstream market, but cloud service providers must be eyeing them with great interest.

Another product that seems SDS-ready is Microsoft’s Windows Storage Server. This is a flexible enough package, that like Ceph, it can reside in an appliance or it can be virtualized onto servers. WSS is richly featured and one wonders if there would be any difficulty in unbundling services to allow a multi-instance flow. This makes sense because of the compute load that services like encryption demand, and the possibility that special hardware assists might be fitted in some servers.

VMWare also is playing in the SDS arena. This is somewhat conflicted, since brilliant success by VMware in this area could facilitate the commoditization that many in EMC must be dreading.

This year also promises to bring some software startups into the SDS arena, both in the US and China. The drive for low cost commoditized hardware is being driven by Chinese ODMs, who have been serving up bare-bones storage gear for several years.

While we won’t see a huge market share shift for software-defined storage this year, we’ll see the beginnings of that shift. The promise of SDS is much lower cost of storage and much more innovation in the way storage is built.

Editor's note: Stay tuned tomorrow for more on software-defined software, when Howard Marks offers his take on this emerging technology.

Learn more about emerging storage trends in the Storage Track at Interop Las Vegas this spring. Don't miss out! Register now for Interop, April 27 to May 1, and receive $200 off.