Even enterprise and midmarket companies, who traditionally have been able to afford to purchase and run their own IT infrastructure, have seen the writing on the wall: it is soon going to be too cost-prohibitive and time consuming to buy and administrate their own on-premise systems. While not everyone is cloud-first, hybrid is starting to gain significant ground.
At the same time, storage requirements are ballooning rapidly. As more devices are connected, more data is collected, and more of business processes go digital, storage needs continue to pile up (plus there’s all that pesky backup data you’ve been holding onto for decades already).
What does the future of enterprise IT storage look like, then? Increasingly, it will be software-defined. Gartner reports that by 2019, 70% of existing storage array products will be available as software-only versions. Software defined storage (SDS) technology enables both object and block/file level storage to be moved across virtualized environments, enabling portability, scalability, vendor agnosticism, and the ability to reuse old or commodity hardware as additional storage.
Software defined storage uses a virtualization layer to abstract available storage locations, separating the control functions from the physical storage. This allows administrators to treat commodity storage blocks or even old servers with internal hard drives as a single, scalable storage pool from which virtual machines can pull resources.
Using inexpensive hardware in this manner can lead to 50% cost savings while avoiding dependence on a particular vendor (at least on the hardware side). Some SDS systems advertise 5x improvements in speed and performance, too.
Storage administration time and costs are reduced, too, thanks to automation features and orchestration improvements. Automation allows SDS to dynamically adjust, like sliding up the storage available to a VM that has a sudden increase in read/write. Orchestration allows admins to set up rules that extend automation, like automatically enabling certain performance settings when a VM is cloned.
If SDS is so great, why hasn’t everyone bought a software layer and applied it to their existing storage? One aspect is that, similar to “cloud computing,” SDS has several somewhat malleable definitions, including traditional storage with a software layer, hyperconverged appliances (which combine storage, networking, and virtualization/compute resources into a single box), and networked storage.
The act of moving systems over to SDS is intimidating in itself, involving training for team members, a possible reduction in staff, and potentially working with new vendors, some of whom may not have a long history.
The best plan might be to take smaller scale projects and implement an SDS solution for them, to get used to the platform and process. Enterprises do seem to taking small steps into SDS, but with some intense interest: one survey found that 26% of respondents planned to implement either hyperconverged infrastructure or SDS within the next year, while a further 38% wanted to deploy it within 18 months.