Since it was first conceived in the mid-1990s, object storage has grown to be one of the largest methods of storing data in use today. The development and commercialization of object storage happened nearly in lockstep with the technology industry that is still its primary user by far, the cloud. Recent advances and shifts in data storage technology have IT buyers wondering if enterprise object storage could become as common as cloud object storage.
Object storage was developed by academic and business researchers to be able to handle what they could see coming down the highway -- a more rapidly increasing amount of digital information that would need to be stored somewhere. So the goal of object storage, right out of the gate, was storage that was less expensive to run and expand. Object storage achieved the first goal by eliminating the need to address every single block of writeable space in the storage medium individually, which is necessary in file and block storage. That allowed more of the medium itself to be used to actually hold data, not information about how to find that data.
The second goal was realized by eliminating any hierarchical structure of folders and treating the storage as a flat address space or a storage pool. This made it easier for the storage to expand as more data made it necessary. Picture an object storage pool as an enormous wading pool that can be added on to in any dimension. The water -- or data, in this metaphor -- simply flows into the expanded space.
Object storage objections
Having a flat address space structure meant it was easier to use inexpensive commodity hardware for object storage, something of great interest to the growing cloud storage business, which was selling itself on being a less expensive way to store your data that isn't in high demand. That has also traditionally been the one drawback of enterprise object storage -- it is slower than file or block storage.
Picture an object storage pool as like an enormous wading pool that can be added on to in any dimension.
A server based on block storage knows where any piece of data is, right down to the individual 4 Kb block of storage media. The level of granularity makes it easier to retrieve a piece of data from the blocks it resides in more rapidly than with an object storage system. That's because an object storage system ignores the individual blocks and treats the piece of data itself as a singular "object." To retrieve it, the storage server needs to find the unique identifier in the entire enterprise object storage pool that has been assigned to the object.
Today, that speed difference has been reduced by advances in enterprise object storage systems to the point that block storage systems like NAS only have the edge in applications that need data available for high-speed transactions. Financial applications like equities or currency trading rely on millisecond differences in data retrieval speed to gain a competitive advantage. But for nearly any other application, object storage has become fast enough.
Hardware and software to the rescue
One of the factors in that speed increase is the growing use of solid-state drives in modern storage systems. As the cost of SSDs continues to drop toward being comparable to hard disk drives on a straight dollar-per-gigabyte basis, more vendors are offering hybrid or all-flash storage. The speed advantage of flash-based SSDs over spinning magnetic disks reduces the speed disadvantage of object storage versus block.
The storage software that manages object storage is also being continuously tweaked and tuned for faster performance, including one of the most popular, Ceph. In April, Red Hat, which provides the open source Ceph commercially, announced improvements to the software that could potentially improve performance eightfold. And, returning to SSDs, Micron announced in May that it was adding support for Red Hat Ceph to its line of SSDs that use the speedy nonvolatile memory express protocol.
One of the last hurdles that enterprise object storage faces is the inertia of changing the legacy way businesses currently handle storage. Most enterprise storage is based on block or file, and is either NAS or SAN. Recently, a few companies have put products on the market that use object storage on the back end, but present to the end user as either NAS or SAN. DataDirect Networks has its Web Object Scaler, or WOS, which can present the storage to the end user as a SAN or NAS, even though it is object storage on the back end.
Exablox has its OneBlox technology, which it describes as object-based NAS. OneBlox is object storage at its heart, but looks to the end user like the type of global file system used in a traditional NAS.
Caringo Inc. has been in the object storage space for years and recently released an update to its Swarm storage software that converts the NFS protocol used in file-based storage systems to its object storage without the need of a separate gateway appliance.
With so many vendors offering products that have object storage at the core and can look like almost any type of storage to the end user, it is clear they understand the value of using object storage for rapid scaling and for cost savings. As even faster forms of nonvolatile memory for storage become available -- like Intel's Optane technology -- even the high-speed applications like financial transactions will at least consider moving to enterprise object storage.
