The distinctions drawn between Infrastructure as a Service and Platform as a Service are getting less clear by the day, with the introduction of DynamoDB by Amazon Web Services (AWS) only serving to help merge the two into a single entity.
DynamoDB is a fully managed NoSQL database service that is designed to provide fast and predictable performance with seamless scalability. Its arrival puts Amazon much closer to the `platform’ territory already occupied by the likes of Salesforce and Netsuite, both of which already provide a rich collection of development tools.
The provision of a database system that can be directly integrated into existing services running on AWS takes beyond its core business so far – IaaS – and marks a shift towards offering additional applications-oriented services that move it to the platform category. It is reasonable to assume that, having made this first move the platform will be filled out with other general purpose applications and development tools, either through further internal development effort or acquisition.
The new database is an internal development and is already in use by many teams and products within Amazon, including the Amazon.com advertising platform, Amazon Cloud Drive, IMDb, and Kindle. The decision to go the `NoSQL’ route also marks what could be the start of a trend. By opting for a non-relational model it is squarely targeting a significant requirement for many cloud-based users the ability to distribute and store high-volume writes in a straightforward and cost-effective manner.
The key issue with relational databases and the cloud is scalability, not just in brute capacity terms but more in the speed and flexibility of scaling and the engineering and capital investment costs of setting up such systems. Cloud-based applications can cause a single relational database to reach its capacity limit rapidly, with the usual solution being to distribute workloads across multiple database servers.
Distributed databases can be complex to set up and manage, and the only alternative to investment in time and skill is to overspecify. Oracle’s Exadata system is a good example of a high capacity, optimised database engine, with the potential downside of a $1 million-plus price tag.
DynamoDB addresses scalability issues by automatically partitioning and re-partitioning data as needed to meet the latency and throughput requirements of applications. There is also pay-as-you-go pricing, so customers pay only for the resources they need.
The end result is expected to offer low, and more importantly predictable, single-digit millisecond read and write latencies at any scale. Data will be stored on Solid State Drives (SSDs) with high availability and data durability provided by synchronous replication across multiple AWS Availability Zones in an AWS Region.
AWS will be offering a free experimental sandbox for developers, with a tier providing 100MB of storage, and five writes and 10 reads per second. That is up to 40 million requests per month free of charge.
The lack of relational capabilities will be overcome through DynamoDB integrating with Elastic MapReduce (Amazon EMR). This will allow complex analytics on large datasets using a hosted pay-as-you-go Apache Hadoop framework on AWS. Using this approach, DynamoDB datasets can be analysed using Amazon EMR, with the results archived Amazon Simple Storage Service (Amazon S3). This allows the original dataset in DynamoDB to be maintained intact.
In addition, it will be possible to use Amazon EMR to access and analyse combined datasets held in multiple stores and store the results in Amazon S3.
Post new Comment