BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
If you're still not sure what software-defined storage is, you're not alone.
Several presenters at the Storage Decisions conference for storage professionals last week in New York gave their take on a software-defined storage definition to help evaluate products sold under that description. The consensus was that software-defined storage is a rolling term tweaked by storage vendors to describe different types of products.
"I don't think the industry has decided what it is. We're just giving different names to things we already had," said Jon Toigo, CEO and managing principal of Toigo Partners International, in his presentation on disaster recovery requirements for software-defined data centers.
Toigo added that software-defined storage applications are merely abstractions of the hardware. Physical storage arrays -- as well as servers and networks -- still host, process and distribute data.
Marc Staimer, president of Dragon Slayer Consulting, said the term is more "marketecture" than architecture, and agreed with Toigo that storage vendors shape the definition to the products they sell.
"This category is so broad and covers so much, it doesn't say anything," Staimer said in his session "Software-Defined Storage … Defined."
Staimer reminded the audience that software has always played a major role in storage. "There is no hardware without software," he said. "Software makes hardware function."
Staimer did offer up the Wikipedia definition of software-defined storage: It's the separation of storage hardware from the software that manages it, and it provides policy management for feature options such as deduplication, replication, thin provisioning, snapshots and backup. It also states that the software is hardware-independent.
Staimer also broke his software-defined storage definition into five product categories: Hypervisor virtual storage, hyper-converged systems, storage virtualization stacks, scale-out virtual storage arrays (VSA) and object storage.
Each category has its own use cases, as well as pros and cons.
Hypervisor virtual storage (VMware Virtual SAN) is designed to solve problems of separate storage silos, high storage costs, complicated storage management and virtual machine (VM) storage performance issues. It gives VM admins control of storage without requiring a separate storage administrator. However, it locks in companies to VMware and has its limits -- 32 clustered hosts, no data reduction, storage tiering or cloud tiering.
Hyper-converged systems combine server, storage, hypervisor and networking in one box. They are aimed at making it easier to integrate, scale and troubleshoot those pieces while reducing the number of vendors needed. However, they can also cause vendor lock-in, raise capital expense in some cases and prevent storage from being used outside of the hyper-converged cluster. Vendors include HP, Nutanix, SimpliVity and VMware EVO: RAIL partners.
Storage virtualization stacks turn x86 servers into storage, often using off-the-shelf JBODs and flash. These stacks are aimed at driving down costs of proprietary hardware and software licenses. They feel like a SAN or NAS and can repurpose older storage, but often require a system integrator. Performance can be lacking with older hardware and they could cause vendor lock-in. Examples of these products include DataCore SANsymphony, EMC ViPR, IBM SVC, Microsoft Windows 2012 R2 and Nexenta Systems NexentaStor.
Scale-out VSAs virtualize hypervisor storage, often pooling storage or flash cache. They are aimed at the same problems as VMware's VSAN but work with different hypervisors. However, they run on VM hosts instead of as part of the hypervisor and can have limited tiering, caching and scaling capabilities. These products include EMC ScaleIO, HP StoreVirtual 4000, Maxta Storage Platform, PernixData FVP and Seanodes Exanodes.
Object storage scales into zettabytes and has the durability to keep data for decades or longer. It is used for geographically distributed data, compliance and to replace NAS for large data stores. Object storage has the greatest scalability among all types of storage and provides built-in disaster recovery and business continuity, but high latency makes it a bad fit for primary storage. Object storage software includes Amplidata Himalaya, Basho Riak, Caringo Swarm, Cleversafe dsNet, DataDirect Networks WOS, EMC Atmos and Elastic Cloud Storage, NetApp StorageGrid, Quantum Lattus, Red Hat Ceph and Scality Ring.
Guide to planning your software-defined architecture
What's true and what's false about software-defined storage options