Performance Object Storage
For global accessibility and Internet-speed application development in web servers, file or image sharing where sub-millisecond latency is not an issue, cloud object storage is fast becoming the platform of choice. However, traditional object storage solutions are slower than many important business applications require for real-time or near real-time analysis, database processing and streaming data delivery.
Because early object storage solutions were intended as secondary storage for archiving, many were built as an additional software layer over hidden file systems and deeper storage architecture stacks, automatically incurring internal performance overhead. For example, OpenStack’s open source Swift object store is intentionally designed to be layerable over other storage (including other object storage solutions) to deliver its repository services.
But today there are new “performance” object storage architectures built from the ground-up to directly manage all the IO directly to disk. This bare-metal object storage can deliver high performance on par with block IO while providing all the other object storage capabilities mentioned previously.
For example, DDN’s Web Object Scaler (WOS) aims to deliver supercomputer-type performance along with the scalability, global access, enhanced data protection and space-efficiency that cloud object storage architectures deliver.
Object Storage Takes Center Stage
Within the next year, we expect system architects to both require and expect object storage as an available platform for newly developed applications, especially web and mobile applications. And IT shops everywhere will be deploying private cloud storage and collaboration projects to keep their users off of public sharing alternatives. Both of these trends will bring object storage from the shadows of archive into the limelight, directly and visibly producing positive business value.
With the evolution of high-performance capabilities, object storage is growing into a primary storage class. Performance object storage can power brand new use cases that could significantly impact revenue opportunities.
One company we’ve recently met is using performance object storage to offer new streaming video applications to large sets of geographically distributed users. These video apps represent all net new revenue streams they could not cost-effectively deploy or operate on file or block based storage systems.
Performance storage could also supercharge fully global collaboration solutions involving large data objects—for example enabling the effective world-wide sharing of large analytic data sets, medical image, or media. We’ve also seen evidence that performance object storage might be a great fit as a persistence layer for larger implementations of in-memory analytic databases like SAP’s HANA.
Object storage has evolved from its archival beginnings to the point where it now makes sense to evaluate it against file and block, especially when architecting new business opportunities. New generations of object storage are not only storing big data, but are powering new kinds of global big data workflows. File and block better watch out because, from our perspective, high-performance high-capacity object storage is clearly primed for center stage.