There are two reasons object storage is considered a good match for the archive storage tier. First, it delivers...
only modest performance and a low cost per gigabyte (GB). In addition, object storage has the capability to extend data durability, and that makes it even more of a natural fit for archiving. But the use cases for object storage are growing. Today's object storage vendors continue to improve the technology and object storage is preparing to move out of its traditional role as an archiving technology -- and into more of a production role in the data center.
To understand the latest trends in object storage, it's helpful to understand why the technology is such a great fit for data not frequently accessed. That data almost always can benefit from the richer metadata attributes that object storage allows. These richer attributes allow for things like multiple versions to be tagged, expiration dates to be set and descriptive context to be applied. All of this makes searches easier to conduct, and the results of those data searches are more relevant with metadata guiding the way.
In addition, object storage systems use either replication or erasure codes for data protection, a scheme better suited for high-capacity hard disk drives (HDDs). The challenge is that object storage's rich metadata capabilities add latency that impacts performance. And since most object storage systems are scale-out in nature, it is difficult for them to use power efficiently; rich metadata requires that all the nodes are powered and available.
So, where does that leave us? The good news is that object storage vendors are addressing these shortcomings. For example, the impact of metadata latency is minimized by storing metadata either in RAM or flash storage. Storing metadata in DRAM -- or on flash -- also enables vendors to address power consumption by powering down off nodes and their HDDs. These improvements also let object storage move beyond its traditional archive role.
Object storage for data protection
A next logical step for object storage is to be a data protection storage area. Like disk backup appliances, object storage systems are very cost effective from a dollar-per-GB basis. With DRAM and flash storage to speed up metadata processing, these systems can ingest data significantly faster than before. In the past, ingestion rates were a considerable challenge to using these systems as backup targets.
While the ingest speed of an object storage system still may not be as fast as a disk backup appliance, they do offer other advantages. The first is scalability. Object storage systems are typically more scalable than most disk-based backup systems. Object storage systems also tend to be more efficient because they can apply deduplication globally. That means the software will compare all files on an array and it will only store the redundant data once. The deduplication occurs across arrays so redundant data won't be stored -- even if those redundancies occur on different arrays. That's a big bonus when it comes to efficiency. Finally, object storage systems are often cost effective since the organization can use commodity, off-the-shelf servers and HDDs.
Now consider snapshots -- object storage is ideal for storing snapshot replication jobs from production storage. An increasing number of production storage vendors (that is, storage vendors who specialize in production storage and not backup and archiving) have direct integration to object storage. This combination allows these systems to keep an updated copy of production data on the object storage system.
Object storage as public cloud replacement?
Another use case object storage vendors provide is as a replacement for file serving. Files are a form of objects, so object storage systems are ideally suited for this task. When combined with file-sync-and-share (FSS) software, these systems can help jettison file serving to the modern era, one where users want to access data from any location and any device.
In this use case for object storage, the speed of ingesting data, or even retrieval, is not a primary concern. Remember that most devices are accessing via broadband or Wi-Fi. So object storage systems are easily able to keep up with those speeds. Reliability and durability are what matters, and object storage delivers.
Most importantly, the combination of FSS and object storage enables IT to eliminate the unauthorized use of public cloud services, commonly known as "shadow IT." Shadow IT puts an organization's data at risk while giving users what they want: on-demand access to data across multiple devices with the ability to share that data with people outside of the organization. In this scenario, a private object storage system combined with FSS can eliminate the use of a public cloud, and that is an appealing option to many companies.
It is true that most organizations can't justify object storage just for FSS. However, when used in combination with traditional archive and backup, the argument for integrating object storage becomes very compelling because the FSS problem can be addressed with almost no additional investment in hardware.
Object storage as a data lake
Another use case object storage vendors are beginning to promote is as a data lake. It's helpful to think about a data lake as an archive, but one that is specifically designed for big data. However, a data lake has to support more protocols than just object storage. Big data is often generated by machines or devices, such as cameras and sensors, which continuously feed small amounts of data to the data center. Typically, a data lake storage repository has to support CIFS and NFS and occasionally iSCSI. The good news is that many object storage vendors have added these protocols. The addition of multiprotocol support makes object storage an ideal data lake because a data lake also needs scalability and durability while being cost effective.
To be clear, object storage will play a key role in data centers to come. In the future, we can expect storage infrastructures to consist of two tiers: an all-flash tier for very active production data, and an object storage tier for secondary data. This means storage pros will be using object storage for archive, backup and file sharing -- essentially any data that won't benefit from flash storage. There are even some all-flash array vendors that are integrating with object storage so that their snapshot replication jobs are sent directly to an object store, providing a more real-time backup and recovery option for this most mission-critical data.
There is no question that object storage is ready to demonstrate how useful it can be beyond archiving, but it is also important to remember that it can fulfill its archive role better than ever. Today's object storage systems have the right protection scheme for high-density drives and are becoming more power efficient. But extending the investment in object storage to do more than just archive allows the system to pay huge dividends.
Top storage vendors offer object storage products
How object storage integrates RESTful API to migrate data
Evaluating open source object storage offerings
Much of object storage use cases point back to the cloud