Amazon Web Services (AWS) offers a variety of storage options, but choosing between them doesn't have to be a complex undertaking.
In fact, experts say, it's increasingly easy to select the right AWS storage service for the right use case. Doing so involves taking into account the needs of object, file or block storage as well as different usage and access styles and requirements.
AWS storage offerings include Amazon Simple Storage Service (Amazon S3), Amazon Elastic Block Store (Amazon EBS), Amazon Glacier and AWS Storage Gateway. The AWS storage options are designed to fit a wide range of workloads, according to Paul Duffy, AWS principal product manager.
Amazon S3 is an object store, which appears much like raw disk capacity. Customers simply create a "bucket"; after doing so, they can put any amount of data in that bucket without deploying anything. "It can be served via Web server instances in the cloud or anywhere else," Duffy said.
Amazon Glacier is targeted more at long-term, archival storage scenarios, typically where access is infrequent. Amazon EBS offers block-store support for server instances, comparable to providing a direct access storage device to a physical server, Duffy said. And, he added, if multiple volumes are attached to the same instance, the user can implement striping and disk snapshots just as they might with physical, on-premises storage. Finally, Amazon Storage Gateway connects the on-premises IT environment with the AWS storage infrastructure.
Weighing the options
Analyst Joe McKendrick recommended that companies start by carefully looking at their workloads, including whether they have intermittent or regular spikes in activity. If so, it may be worthwhile to consider a solution that is, in his words, "there when you need it, and you are only charged for those instances when you use it." AWS isn't the only provider offering a way to handle "bursty" workloads, he notes; RackSpace, Microsoft and IBM, among others, also offer on-demand cloud services.
It's important to understand the difference between object storage and file or block storage, said Arun Taneja, principal analyst at the Taneja Group. Cloud storage services are generally object-based, which simplifies the task of getting data to the cloud by leveraging the representational state transfer (REST) application program interface (API). "That's the good news," Taneja said. "The bad news is that nothing on the enterprise side is built that way."
For that reason, organizations with established storage patterns and resources that want to integrate AWS storage will have to consider rewriting applications with the REST API or using a gateway product -- potentially including AWS Gateway, Taneja said. A gateway can make the AWS storage look like a familiar interface such as the Internet Small Computer System Interface (iSCSI). Longer-term, Taneja said REST will probably be built into applications from the start, making adoption of AWS and similar services even easier.
With those interface details out of the way, almost any application can start to use AWS and get "reasonable" performance, he said. Still, he noted, due to potential latency issues, many applications will still benefit from being routed through an intermediary such as a gateway service.
AWS customers can choose the right storage service for the right use case, according to Myles Weber, vice president for cloud and community services at Appian, a business process management software vendor that uses AWS for most of its IT functions. "If you need to store lots of data but only access it infrequently, AWS has a very inexpensive but somewhat slow offering. But if you need lots of data with faster speed and frequent access they can do that, too," Weber said. "For a company like us that is focused on innovation, [AWS makes] a good partner because they can easily accommodate our 50% growth."
However, Weber added, while AWS storage is cost-competitive, potential customers should understand exactly how charges are accrued. For example, one key measure, Input/Output Operations per Second, can contribute to charges, "but only a few people really understand how that works," he said.
Similarly, whether it is AWS or another provider, it's important to examine the contract terms closely to see how data is treated, McKendrick said. Most vendors word their contracts to indicate that they aren't liable for lost data, so customers should be sure they're backing up their data elsewhere.