BIG DATA: the internet: home to Big data innovation ================================= Not surprisingly, most of these game-changing technologies were born on the Internet, where Big Data volumes collided with a host of seemingly impossible constraints, including the need to support: • Massive and impossible to predict traffic. • A 99.999% availability rate . • Sub-second responsiveness. • Sub-penny per-session costs. • 2-month innovation road maps. To satisfy these imposing requirements constraints, Web entrepreneurs developed data management systems that achieved supercomputer power at bargain-basement cost by distributing computing tasks in parallel across large clusters of commodity servers. They also gained crucial agility – and further ramp-ed up performance – by developing data models that were far more flexible than those of conventional RDBMS. The best known of these Web-derived technologies are non-relational databases (called “NoSQL” for “Not-Only-SQL,” SQL being the standard language for querying and managing RDBMS), like the Hadoop framework (inspired by Google; developed and open-sourced to Apache by Yahoo!) and Cassandra (Facebook), and search engine platforms, like CloudView (EXALEAD) and Nutch (Apache). Another class of solutions, for which we appropriate (and expand) the “NewSQL” label coined by Matthew Aslett, strives to meet Big Data needs without abandoning the core relational database model.4 To boost performance and agility, these systems employ strategies inspired by the Internet veterans (like massive distributed scaling, in-memory processing and more flexible, NoSQL-inspired data models), or they employ strategies grown closer to (RDBMS) home, like in-memory architectures and in-database analytic s. In addition, a new subset of such systems has emerged over the latter half of 2011 that goes one step further in physically combining high performance RDBMS systems with NoSQL and/or search platforms to produce integrated hardware/software applications for deep analytic s on integrated structured and unstructured data. to be contd . . .
Posted on: Wed, 09 Oct 2013 03:41:24 +0000
Trending Topics
Recently Viewed Topics
© 2015