PostgreSQL 9.2 released

QArea Expert by QArea Expert on September 13, 2012

PostgreSQL 9.2 released
Reading Time: 2 minutes

On September 10, 2012 the PostgreSQL Global Development Group officially announced PostgreSQL 9.2 release!

Finally the long period of beta testing declared in May of this year was completed. Developers promise the improved performance, scalability and flexibility.

The improved performance and scalability

Now Postgres supports 64-nuclear processors and its energy use was decreased. Everything is clear and without superfluous retreats.

The index-only scan mechanism is realized. The matter is that Postgres’s indexes don’t contain any information about visibility of record. Thus to understand if the record is visible the server needs to read a real tabular tuple. There is a situation that the index contains information about out-of-date tuple. The access to the sorted index is much faster than access to the real table. There is no information about visibility of record but developers realized so-called “visibility map”. But the visibility map has been created earlier. In the new release it was made steady against server falling as well as guaranteed correct. Also Streaming Replicatio is improved and polished.

There was created a JSON type parent support. The server checks a validity of entering data and shows defects. It also contains a full set of functions for converting in JSON.

And the greatest feature became the support of Range Types:

  • Ranges can be both discrete and continuous. The discrete types are int4range, int8range and date4range. The continuous types are numrange, tsrange, tstzrange (ts — timestamp, num — numeric).
  • The range can be opened. Probably you know about round and square brackets: [1; 1] — closed, and [1; 1) — opened in the right side etc.
  • The infinity (∞) can be border of a range!

Earlier this feature we had to use two fields for storage of such data. And now we have operators, indexes and constraints.

Such comparative numbers with the last versions are given in the official text:

  • About 350000 requests for reading per second (faster in 4 times);
  • Index-only scannin can give a gain in speed from 2 to 20 times;
  • About 14000 data trackers per second.