MariaDB Vision Majorca 2009
From Askmonty.org
(Redirected from MariaDB Vision Majorca2009)
This page presents ideas for the future strategy and technical direction of the MariaDB product. Presented by 3 groups at the Monty Program Staff Meeting in Mallorca 21.Aug 2009. This will be used as input when defining suggestions for the development roadmap.
Contents |
Group Bryan
The Future
- SSDs, many many cores (without individual core performance), mobile computing, distributed computing
Strategy
- "SSD Mode": Implement optimizer and other improvements for SSD, seek operations are no longer expensive
- New Parser: Improve performance considerably for fast queries (majority of websites)
- Modularity: Make it easy to only include what you need (good for mobile computing) or add new special features.
- Implement features users actually want and need, keep current on what would help people the most. Possibly scan public MySQL worklog for "low hanging fruit", needs reported by users/customers
- Improved query cache (or caching in general)
- Improved GIS (important for mobile computing)
Group Hakan
Trends
- Running a single MySQL server on a single system is going away.
- Data is growing all the time.
- People are taking more and more digital video and need a way to store and manage it.
- Multimedia and graphic artists are using databases to run large video installations.
- Server hardware is multicore with large amounts of storage and RAM
What is the right strategy for the MariaDB product?
- There are some aspects where MySQL is great, but for other things it is very poor. Postgres, by example is very balanced, things are pretty much the same throughout the server. MariaDB needs to be more balanced.
- Should we compete with Sun on hardware-level and engine-level performance? We don't have the resources, so maybe not.
- A good ways for us to compete would be to optimize at the SQL level.
- We could also focus on being friendlier to outside contributions.
- Stability & Reliability are very important to end users.
- Resume building MariaDB from the plans that were in place 2006 to make it more mature.
- Should we kill the parser and create a new one? The parser sucks a lot of CPU.
- Be the default and best database for PHP/Python/Perl
- Figure out what Web/Web 2.0 means and create a meaningful benchmark
Group Bo
Databases in 5-10 years
- Move towards primitive distributed key/value storage
- Ajax: Many small hard-to-optimize queries -> automatic indexing or caching
- Massive amount of data
- More distributed installations
- GIS
Technologies
- Dynamically typed databases - for dynamically typed languages
- Getting rid of limitations like varchar lengths
- Hiding SQL from the programmer - intent based programming
- Link
- Hibernate
Hardware
- SSD
- Consumer devices (GIS, telephones, cameras, amarok/mysql)
- Multicore - cloud
Stable future
- More regression testing
- Performance regression testing
- Code cleanups - lower barrier to entry
- Add diagnostics
Sweet spots
- Massive community patch acceptance would change the world
- HTML 5 drop-in database
- Involved in standardization?
Strategy
- Improve patch submission:
- Stable API internally
- Code cleanup
Current problems: We want to include many patches but focus on stability. There is a steep learning curve. We need to learn how to accept patches.
- Improve multicore support
- Improve code quality
- Add diagnostics
- Use storage engines to handle SSD
Further Suggestions
- For users with large server groups Manageability features are very important
- Need for these features will grow
- Performance Interface has been requested by users with large installations already for years
- Synchronization between devices
