| |
| META TOPICPARENT |
name="LargeScaleDataManagement" |
NoSQL Systems
The increasing data volume that needs to be stored, indexed and queried for every organization (such as e-mail and web logs, historical data, click streams, etc) has pushed classic database systems to their limits. The weaknesses of classic databases to deal with large scale data analysis tasks [1], along with the embarrassingly parallel nature of these tasks, has lead to the development of horizontal scalable, distributed, non-relational data stores, called NoSQL databases [2]. Google's Bigtable [3], Amazon's Dynamo [4], Facebook's Cassandra [5], and LinkedIn? 's Voldermort [6] are a representative sample of such systems. In favor of scalability and high availability, NoSQL systems relax typical ACID guarantees made by typical DBMSs, allowing, for instance, only eventual consistency. NoSQL systems can serve a dual purpose: they can efficiently store and index arbitrarily big data sizes while enabling a large amount of concurrent user requests. Recently, the cloud computing paradigm is increasingly gaining attention both from the industry and academia. "On demand" and "pay as you go" access to computational and storage resources that reside in distant data centers is a very attractive business model, especially for small companies or start ups that need a quick, cheap and scalable access to hardware and software infrastructure. NoSQL systems are perfect candidates for cloud infrastructures, as their "shared nothing" architecture enables them to scale by simply acquiring more computational and storage resources from a cloud vendor. |
