Managing data with an object storage system
A comprehensive collection of articles, videos and more, hand-picked by our editors
Object-based storage is considered a leading emerging technology, and many of its features are deemed ideal for cloud storage. However, drawbacks such as lack of standardization and performance limitations often relegate object storage to archiving and backup, and even then it can only be used with cloud gateways.
Benefits of object storage include massive scalability, geographic independence, multi-tenant features and the ability to use off-the-shelf hardware. Disadvantages include poor IOPS and high latency, but it's important to keep in mind that each vendor’s technology is proprietary. These shortcomings make object-based storage less useful for primary storage.
Most enterprise applications for unstructured data were designed to speak to traditional file system and block-based protocols, such as iSCSI, Fibre Channel (FC), Network File System (NFS) and Common Internet File System (CIFS). Object stores, however, can only speak natively to applications that have a Web-services interface and many legacy applications haven't been architected to speak natively to object stores.
“The way people implement cloud is generally through object stores,” said Robert Mason, Nasuni's president and founder. “You need to do a customized interface to the object store, [but] there's no API standard for either private or public object stores.”
Lack of standardized cloud storage application programming interfaces (APIs) means customers will have to deploy cloud storage gateways as the protocol translators to avoid the arduous task of rewriting applications to support Web service interfaces.
When the Park City, Utah-based Sundance Institute explored using cloud APIs to store and retrieve high-definition video files and photos for a public cloud implementation, its IT staff realized the institute’s press.sundance.org application would have to be substantially rewritten to talk to cloud providers’ APIs.
To clear that hurdle, Sundance purchased a StorSimple iSCSI cloud gateway for its hybrid cloud to avoid having to “mess with any APIs,” according to Justin Simmons, associate director of systems and IT administration.
“[Using cloud APIs] would require that I customize the application to work with the cloud provider’s APIs, and if I wanted to work with another cloud provider I’d have to rewrite the application,” Simmons said. “The gateway solved the problem for us so I don’t have to involve the developers.”
Cloud gateways to the rescue?
The lack of standardized cloud storage APIs has made gateways a crucial component for cloud adoption, largely because the devices work as the translators between applications and Web services-based APIs that include the HTTP-based REST, Amazon Simple Storage Service (S3) APIs, EMC Atmos APIs, Nirvanix API, Rackspace Cloud Files APIs, and the Simple Cloud API along with the Storage Networking Industry Association (SNIA) Cloud Data Management Interface (CDMI).
Vendors such as CTERA, Nasuni, Panzura, Riverbed, StorSimple, TwinStrata and Zetta have developed cloud gateways either as hardware appliances or software that customers can install on their on-premises devices. Public cloud storage providers are also adding gateways that make it easier for customers to connect to their clouds.
In January, Amazon announced its software-based AWS Storage Gateway. Nirvanix and Rackspace also offer file-based gateways, and industry insiders expect Windows Microsoft Azure to add a gateway as well.
“I get a lot of questions from customers about gateways. [Gateways] are solving a lot of problems,” said Gene Ruth, research director at Stamford, Conn.-based Gartner Inc. “But the trouble with gateways is they're small scale. They're good for the SMBs [small- and medium-sized businesses] for small-scale requirements.”
Jérôme Lecat, chief executive officer (CEO) at Scality, agreed, and said gateways are good enough on a small scale “but they're stretched on a large scale.” Across the board, analysts say gateways meet the basic needs of cloud customers but they rarely hold up in the enterprise, which is why they're most often used for backup and archiving.
“The problem with a gateway is it adds more complexity and limits performance,” said Andrew Reichman, principal analyst at Cambridge, Mass.-based Forrester Research Inc. “I think from a performance and elegance perspective, it’s not ideal, especially for primary applications.”
Organizations are testing cloud storage by moving backup and archiving applications to object-based cloud storage. This is because these applications are more write-intensive, tolerate low latency and can be more easily redesigned to talk to object stores. It’s more difficult to rewrite applications to speak to object stores that are read- and write-intensive
“Object stores are powerful, but they're raw,” said Andres Rodriguez, Nasuni's founder and CEO. “Object stores fundamental operations are PUT, GET and DELETE, while block and file allows for modifications. The ability for modifications in object stores is difficult. Backup and archiving applications have no modifications, so it’s a natural use case for object stores.”
For large data sets, object-based storage devices offer alternative to file systems
Object storage from Caringo CAStor helps Maine-based Synergent Corp.
Spring 2011 Storage Networking World news: Object storage system by Amplidata launches