Announcement

Collapse
No announcement yet.

PostgreSQL 14.0 Released With More Performance Improvements

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by elatllat View Post


    MariaDB and CockroachDB have features some PostgreSQL users have wanted for a long long time.
    CockroachDB literally uses the postgresql database engine, just backed by concensus replicated writes across multiple replicas.

    Unless "true multi master" and "scales beyond a single node" are the features you were talking about, I'm not sure how that's possible

    Comment


    • #12
      Originally posted by partcyborg View Post

      CockroachDB literally uses the postgresql database engine, just backed by concensus replicated writes across multiple replicas.

      Unless "true multi master" and "scales beyond a single node" are the features you were talking about, I'm not sure how that's possible
      Yes. CockroachDB lacks blob support though so Ceph may also be used.

      Comment


      • #13
        Originally posted by uid313 View Post
        Great database!
        ...And create a single program to be the entry point of all functions just like Git does with the "git" command.
        No program needed, a simple script will do.

        Comment


        • #14
          Originally posted by bug77 View Post

          No program needed, a simple script will do.
          Yeah, a script that is a front-end to all pg_ programs would be great, then hide all the pg_ programs somewhere where they are not in the $PATH.

          The PostgreSQL daemon might be great but the user space could be better.

          Comment


          • #15
            Meanwhile Embarcadero is making a major push to market their Interbase database....

            It just achieved recursion for its Common Table Expressions yesterday; PostgreSQL had CTEs with recursion 11 years ago. It took EMBT 4.5 years after adding CTEs to implement the recursion functionality. It still touts "SQL-92 compliance", has no partial indexes, no window function support, no JSON support, no XML support, no full text search, no GIS support, no timezone support, no UPSERT, and just added tablespaces support (PostgreSQL had that 15 years previously). A server version with unlimited users costs $7,000 USD for 8 cores with $1000 for each additional 8 cores; 32 core maximum. You can get a Windows-only desktop version for $60 that supports 4 cores, 1 user, and 20 connections. It has an embedded version as well; no one knows the cost per user to embed it in your software.

            They delete/don't approve comments on their blog pages that ask about the lack of basic SQL features like window functions or ask about comparisons to PostgreSQL (their product comparison page omits PostgreSQL). I found an old brochure from more than ten years ago that mentioned PostgreSQL; their arguments were insane. They complained about all of the adjustable options PostgreSQL has! They like to tout how their database is "zero administration"; my car is self-driving too if I put a brick on the accelerator pedal; of course it'll just crash into something or drift into a ditch. For instance, they don't have anything like PostgreSQL's auto-vacuum functionality - if you delete a million records it does so instantly, but it sweeps the database with the next query so a simple select statement might need to wait for those previously deleted million records to be purged. The only other option if full manual. Their other complaint was that PostgreSQL produces a lot of logs and those logs will have to eventually be compressed or deleted! Um... this is called "server log rotation" and happens on every server, often automatically. These were the best criticisms they could come up with!

            In their current comparison to SQLite they make reference to unspecified disaster recovery and high availability functions of Interbase (as if you need multi-version concurrency control on a phone) and conclude that SQLite "isn't really suitable for ISV development"! Meanwhile, SQLite is the most-used database on Earth, in every Android and iOS device, OS X, WIndows 10, every major browser, etc.

            These are strange times in which closed-source software vendors seem to be getting desperate and have picked up some tips from Trump....

            Me, I'm happy with the combination of SQLite and PostgreSQL as the ultimate embedded/server-desktop database duo.

            Comment


            • #16
              Originally posted by partcyborg View Post

              CockroachDB literally uses the postgresql database engine, just backed by concensus replicated writes across multiple replicas.

              Unless "true multi master" and "scales beyond a single node" are the features you were talking about, I'm not sure how that's possible
              Nope, Cockroach is just (mostly) wire-compatible with Postgres but does not share any meaningful code with it: https://github.com/cockroachdb/cockroach
              Not sure what you meant by a database engine, but the closest thing to it would be its underlying KV storage, RocksDB/Pebble.

              However yeah, when it comes to features it doesn't bring that much to the table besides scalability/replication, their SQL implementation still lacks some (not so commonly used, tbh) features that Postgres has had for years like e.g. FTS or RANGE types.

              Comment


              • #17
                I have not heard about cockroach before the comments above, but in comparison to PostgreSQL galera cluster build atop of mariadb gives true multi-master SQL replication, i.e. if you take 3 computers and make a cluster and put these computers in 3 different data centers it works like a charm and if any of them fail the database goes on to operate normally without end-users even to notice anything. Postres basically cannot do this. It has Master-Slave replication but it is problematic there is a period when database is not working, then you have to write your software in such a way it would be switched to a slave, then you have to think "who is the master today" so on. What is interesting initially I tried to use official MySQL multi-master replication and it works worse than galera. The genuine mysql multi master cluster can become unstable and fall apart without any obvious reason and it's super hard to build it back.

                Comment

                Working...
                X