The future of DBMS without SQL

"Two-level" memory when DBMS at the appeal to a storage system uses not only own, but also internal memory of DWH is necessary for speeding up. At the same time some DWH actively uses memory not only for reading, but also for record (with background record on disks). For speeding up some vendors offer in-memory-SUBD or similar technologies when on a disk only the log is written (consistently), and all data while they are located, are in memory.^[1]

Reliability of a system consists of two components. First, it is necessary to secure itself against data loss which can happen because of failure in DWH, on the DB servers or because of errors in software. To avoid data loss, it is necessary to do mirrors, to keep the magazine of changes, to create backup copies of bases on separate storage systems and then to transfer these backups to tapes.

It is worth to remember that in addition to protection against data loss it is necessary to ensure still and smooth operation of the DBMS. Most vendors offers these or those forms of cluster solutions for minimization of idle time in case of failure of DBMS now.

Besides, become very popular Appliances – boxed solutions. Actually it is return to "hardware-software complexes". All vendors have such solutions practically, and they not only are balanced, but also can brag of lack of "single point of failure".

The market of DBMS promptly develops, and the future belongs to Appliances and MPP (massive parallel processing) to systems which easily extend, and their performance linearly increases when adding new servers. Such solutions in this or that form appear at all main vendors. Among them systems allowing to build own configurations are optimal, using the recommendations of DBMS producers. It is normal cheaper than Appliances and, besides, is significantly more flexible at other limited resources – for example, at the limited place or restrictions on loading or cooling in DPC.

Use of MPP technologies goes side by side with transfer of databases to clouds. Reliability and high-speed performance is provided with the correct configuration of such systems. As a rule, "cloud" providers are engaged in it. In that case there is a question of confidential data protection. The most important point is a trust between users of cloud services and provider here though there is also a problem of protection of communication channels between "cloud" and users of a system. However you should not dramatize a problem of confidential data as there is no consensus what data actually should be closed.

At one of conferences of CNews on the subject "Big Data" interesting examples from practice of the British government which refused to open data were given, but officials after all forced to publish information after long legal procedures. When data appeared in broad access, it became clear that in government base there were wrong data on about 16 thousand from three hundred with something thousands of bus stops. And thanks to the fact that data on schools were published there were enthusiasts who fastened to them the web interface and free of charge made for the government the mass of analytics. Providing free access to this information allowed not only to modify inaccuracies, but also to process data without increase in the budget.

One more trend of the market is a development of DBMS with the open code which do not concede in the speed and reliability to products of IBM, Microsoft or Oracle. The NoSQL technology enjoys special popularity. It is necessary to emphasize that recently just vendors bethought and began to be implemented into the market of MPP and NoSQL actively. Oracle released ExaData, Microsoft purchased DatAllegro written on the basis of open Ingres DBMS and began to release the version of MPP Appliance, IBM purchased MPP Appliance Netezza using open DBMS Postgres, EMC purchased MPP software Greenplum which not only uses Postgres, but also actively develops it. Besides, all leading vendors now actively enter or already Apache Hadoop entered the NoSQL market with the implementation. So still open DBMS advance vendors and move this industry forward.

Notes