MariaDB roadmap session

Discussion of things to possibly include in 10.2

Transcript

1. MariaDB 10.2
   Roadmap
   Rasmus Johansson
   

   View Slide

2. Pouring and flooding
   15/10/2015 2
   Mission 10.2 - Squad Shot
   Those stupid bugs think they
   can take us out with superior
   numbers. They should think
   again.
   https://www.flickr.com/photos/masked-builder/6176765530
   

   View Slide

3. Windowing functions
   • MDEV-6115, Window
   Functions as in the SQL
   standard
   • MDEV-8091, Simple
   Window Functions
   15/10/2015 4
   • Parser work close to finished
   • Looking at executions side
   • Look into re-using test cases from
   PostgreSQL
   • EXPLAIN JSON support
   

   View Slide

4. JSON
   • Virtual columns with functional indexes (i.e. generated columns like in 5.7)
   can be implemented within a couple of weeks
   • MySQL 5.7 uses non-standard function names
   • Suggestion use standard names when available, create UDF to match
   MySQL names
   • Come up with Usable JSON Test suite
   • Maybe more strictly typed than MySQL implementation
   • Possibility to choose between lax and or strict mode
   15/10/2015 5
   

   View Slide

5. InnoDB in 5.7
   Small changes in the interfaces
   • New durability declaration
   • New index pushdown interface
   • New thread interface
   • New thr cond, lock, mutex, rwlock interface
   • PSI interface changes
   • PFS interface changes
   • New XA interface
   • New THD functions
   • New class functions (ha_innobase, API change ?)
   

   View Slide

6. InnoDB in 5.7 continued
   • Changes on tablespaces
   • CREATE TEMPORARY TABLE ... DATA DIRECTORY={path}
   • CREATE TABLE ... DATA DIRECTORY={path} TABLESPACE={name}...
   • MERGE_THRESHOLD= // New SQL
   • ALTER TABLESPACE // New SQL-clause
   • New structures for create table/alter tablespace
   • GIS indexes
   • Index concurrency improvements
   

   View Slide

7. Related in 5.7
   • Performance Schema changes
   • Virtual Indexes
   • Indexes on non-persistent virtual columns
   • Optimizer pattern recognition to match
   expression in the WHERE clause against indexes in
   virtual columns
   

   View Slide

8. Replication
   • Semi-sync replication
   • Galera 4.0 + other improvements
   • Critical cloud deployment
   • Google patches (GTID indexing, “no slave left behind”)
   • Tools for automating failover (Shlomi’s orchestrator)
   • Crash-safety with log position
   • MariaDB replication in Performance Schema
   • Rollback binlog to certain time in history
   15/10/2015 9
   

   View Slide

9. Encryption
   • Support for key management
   systems
   • Support for backup systems
   10
   

   View Slide

10. Security
    • Kerberos plugin (maybe for 10.1)
    • Expiration of user passwords (MDEV-7597)
    • Block user accounts after failed login attempts
    (MDEV-7598)
    • Encryption enhancements
    • Key management
    

    View Slide

11. Packaging
    • Better documentation for build packages on your
    own
    • Interoperability between MySQL, MariaDB and
    Percona server
    • Configuration and data separation
    • Use one packaging system to generate all packages
    • Can Cpack be used for all packages?
    

    View Slide

12. Storage Engines
    • Spider
    • RocksDB
    • ScaleDB
    Storage Engines
    Change to API:
    • Query fragment push down, server side part and then
    another part in the storage engines that want to use it.
    • Remote storage engines like FederatedX, Connect, Spider
    will get performance boost
    • https://dev.mysql.com/worklog/task/?id=4292
    

    View Slide

13. Storage Engines
    • Spider
    • RocksDB
    • ScaleDB
    Storage Engines
    • Spider
    • Cassandra (disable from compiling)
    

    View Slide

14. Connector requirements for server
    1. Binary protocol / Prepared Statements
    a) Batch update
    b) Prepare & Execute (prepare the placeholders for the
    values (?) )
    c) Option to return resultset only once
    (now returned twice or more)
    2. Support for native Windows SSL
    (SChannel)
    

    View Slide

15. MaxScale issues for server
    • A user has to be created twice
    • Once for her actual host: ‘user’@’user_host’
    • Once for the Maxscale host:
    ‘user’@’maxscale_host’
    • Grants have to be given to both identities.
    • Double parsing of queries
    • Passing of parsetree to server
    

    View Slide

16. View Slide