My Coupon Database
  • Important developments in the field of computer hardware ongoing


  • The ability to build relatively cheap, unlimitedly horizontally scalable cluster systems (whether public or private cloud infrastructure systems or clusters configured in a traditional way) has led to a dramatic increase in research and development of data management system architectures suitable for such environments without sharing resources (shared nothing). Work is under way on two main fronts. (I won't talk here about companies producing basic SQL-oriented DBMS, which always try to solve all the problems at the expense of their own, accumulated over decades of capabilities).Important developments in the field of computer hardware ongoing
    The first front are the main Internet service providers, such as Google, Yahoo!, Facebook, Amazon, etc., which for their own needs and to provide publicly available cloud services develop data management tools that ideologically and architecturally move away from the traditional canons of the database community. In many ways, it is associated with the activities of these companies associated with the emergence of the concept of "NoSQL", ie (in its original sense) the rejection of existing solutions.
    On the second front, in my opinion, are the followers of Michael Stonebraker's concept of "one size is unsuitable for all", in which, in fact, two considerations are basic: (a) the time has passed for universal database management systems suitable for supporting all kinds of applications, and (b) the development of new systems must take advantage of all the useful technologies and ideas accumulated in the database community. This camp includes research teams from several US universities, VoltDB, Vertica, Asterdata, etc.
    It should be noted that the representatives of both fronts agree on the first point, i.e. they are united by the desire to create new data management tools that are specialized for some classes of applications and support these applications more effectively than universal DBMS. The second point divides them: people from the first camp think (a little bit exaggerated) that database technologies are a heavy burden of the past, and the second camp treats them as a valuable legacy that cannot be sacrificed.Important developments in the field of computer hardware ongoing
    On both fronts there are works in two most important directions of data management - analytical systems of data management (i.e. systems suitable for building OLAP category applications (online analytical processing, operational analytical data processing) and transactional systems of data management (i.e. systems suitable for building OLTP category configuration management tools applications (online transaction processing). In the first direction the representatives of the two camps under consideration in the past years were divided, first of all, by the attitude to NoSQL technology of data analysis MapReduce. Not so long ago, representatives of the NoSQL camp believed that MapReduce will replace the performance tuning in oracle parallel analytical systems of databases in dynamic cluster architectures, and researchers from the second camp accused the creators of MapReduce to return to the dreary past, when the database technology did not exist. However, this time seems to have passed. At least the database community accepted the MapReduce technology and learned to use it,and I (maybe temporarily) consider this topic closed.