什么是软件定义存储 (SDS)？

Software defined storage (SDS), like software defined networks, is software used to create pools of storage composed of multiple storage devices which then appears as a single unit, with the goals of adding flexibility, diversification, and enhanced functionality of the underlying storage. In short, SDS abstracts the storage management software layer away from the hardware device, effectively emancipating these management functions from the confines of each device being managed.

SDS is particularly useful in dynamically changing circumstances. In particular, SDS makes it possible for organizations to source storage from multiple vendors and data storage types that help achieve resilience and improve data availability. With SDS software, these multiple sources can be automatically provisioned, de-provisioned, and the whole can be scaled to meet data demands by simply adding storage to the software defined system.

SDS vendors should add features that include the ability to interact efficiently and effectively between all the underlying storage components. The Storage Networking Industry Association (SNIA) defines five characteristics of software defined storage.

• Automation – Automation helps to simplify management, reduce costs, and maintain storage operations between the diverse storage make-up underpinning the software defined storage system.
• Scalability – While many systems can scale, SDS must be seamless so that storage appears infinite in theory, when capacity must increase, it must do so without hiccup.
• Standardized Interfaces – SDS must supply APIs that facilitate automations, and ultimately management, maintenance and provisioning of data storage.
• Transparency – The SDS system must allow for monitoring of storage consumption so that it can be easily managed, and can inform higher level business concerns like cost.
• Virtualized Data Path – Virtualized data paths allow block, file storage, and object storage to interface between applications.

SDS vendors can fulfill both customized and turnkey software defined storage solutions. Within these two needs, several degrees of SDS can be found, so it is advisable to compare product offerings. These are not hardline categories, but rather ones that demonstrate how SDS technology is evolving.

• SDS Software Only — The basic solution is software only packages that allow for virtualization, and optimization of underlying storage devices.
• Software Layered on Enterprise Storage Arrays — Sometimes referred to as scale-out network attached storage (NAS), these hardware devices can be grouped into logical units or clusters. They are typically geared for performance of unstructured data, and specialized to bring costs down. These may also come in the form of appliances, and are marketed as hyperconverged infrastructure (HCI).
• Cloud-based SDS — Becoming more the norm for data storage are cloud-based vendors of SDS software which requires little setup from the purchaser. These HCI tout their ability to support public, private, and hybrid cloud environments with the greatest performance.
• SDS with AI Capabilities — Found in other packages, like cloud-based SDS, are more advanced AI features. AI helps admins analyze performance and space use, which translates into understanding when capacity needs to be increased, where bottlenecks may be dampening performance, and when it's time to decommission older storage devices.
• Hyperconverged Infrastructures — While SDS is often marketed as Hyperconverged Infrastructure, HCIs do more by consolidating the advantages of virtualization, servers, storage, and networking into one package reducing complexity and subsequent administration, while eliminating compatibility issues. In comparison, SDS can be thought of as an element within an HCI ecosystem, whereas HCI is a central consolidated point of control over the entire infrastructure.

Businesses benefit in key ways from the storage abstraction provided by software-defined storage platforms.

• Ultimate Choice in Underlying Hardware — Software-defined solutions are inherently designed to be agnostic. Consequently, organizations can include any variety of vendor storage in their storage infrastructures, all managed from a central point.
• Centralized Management — SDS software is uniquely designed to make visible all storage aspects and connections between underlying storage infrastructure. This helps to clarify the entire storage system, and simplify management.
• Reduced Storage Management Costs — Storage visibility brings into focus unused space which with storage optimizations can maximize system performance. With SDS optimizations, costs can be controlled and businesses can take action to reduce expenses.
• Automation of Key Storage Management Functions — Routine and error-prone tasks that were once the responsibility of human administrators can be shifted onto automated management functions. Automated functions include storage optimizations, replication, data migrations, and backups.
• Unlimited Capacity to Scale — Because an SDS management layer is decoupled from the hardware storage, storage can be added and removed from the infrastructure pool dynamically. Theoretically, the system can scale infinitely. This beneficially also allows for aging storage to be added to the pool, gaining the functionality of the SDS management layer, and repurposing unused storage assets.

Software-defined Storage (SDS) solutions serve multiple use cases.

• Extended Storage Network Configurations — In the instance of remote office branch office (ROBO) configurations, SDS helps to tie branch storage together with the central infrastructure. Also, SDS is a cost-effective solution for aging branch infrastructure, rather than choosing the costly alternative of replacing existing structure with usable life left.
• Rapid Response Situations — Emergency situations may call for the deployment of multiple storage units within an emergency operational area, SDSes can rapidly aggregate these deployments when time is critical.
• Hybrid Cloud and Infrastructure Modernization — SDS can be used as a bridge for companies extending their private clouds into hybrid clouds. As well, the same bridging function can be used to help companies modernize their storage infrastructure, allowing them an easy means for cloud migration.