Whim, for a database, we think of some interesting things and functions, be regarded as a kind of future database development Imagine it, look forward to the future strength of the company or research institution can really achieve.
A: distributed operating system --- perpetual motion machine type operating system
Chat database, first talk about the operating system, after all databases to install and deploy the operating system.
Hard fault current common application services into power, CPU, hard disk, memory, network, etc., which are common, irresistible fault. Under non-distributed operating system, not the individual server resources to maximize the rational use, even now K8S container era, as well. Currently the application service level, there is read and write performance, CPU, memory, network bottlenecks, but the database in a distributed operating system will break through these bottlenecks, the type of application to achieve perpetual motion, never down, sufficient resources case, automatic resilient and elastic resources, performance will never be the bottleneck, the overall use of resources reasonably, application services are no longer concerned about hardware problems.
Currently a distributed (such as ceph file system, etc.) to store files, but not yet implemented a distributed CPU, memory, network, even if only to realize a distributed file system, but also brought great changes in the database, representing Aurora, PolarDB etc., to achieve separation of storage and computing, hardware, basically do not have the concerns of hard drive failure, safety, performance and elastic scalability have been greatly improved.
If the main CPU, hard disk, memory, network, all to achieve a truly distributed, no longer distinguish between single CPU, hard disk, memory, network, all the resources are in a big pond, all distributed cluster CPU and memory There is a cluster of other resources can be shared, either limited use service resources can be automatically extended elasticity, so it is possible to achieve 100% reliable application service, then service is not running on a physical machine, is a dynamic run in a distributed operating system, even if one physical node failure, will not affect the normal operation of the business, it does not affect the safety and reliability of the data.
Cost is not left to talk about architecture and practical, the emergence of such a distributed operating system, hardware and software support needs, but with the rapid development of software and hardware technology, to believe in these operating systems in the future.
II: intelligent database
The current database intelligence, it should be said that automation, intelligent operation and maintenance are mostly true, not many truly intelligent company. For intelligent database, individuals should Imagination is the use of machine learning, automatic, adaptive complex environments, unattended installation deployment, operation and maintenance, and database and keep the best of the best run state.
1: Automatic Performance Tuning, no profile of, CPU, memory, hard disk, network, kernel, thread, TCP connections automatic tuning, database startup operation start to keep the setting of various parameters is optimal , database performance is optimal.
2: automatically add and index optimization, database tables do not need to specify the index when it is created, to create the optimal index by the database based on real situations.
3: The database operation and maintenance personnel can dynamically specify development standards, such as disabling a large field, there is no primary key table directly in the creation of an error, increase auditing database development specification.
4: automatic rewriting SQL statements for complex SQL can automatically rewrite, simplify and optimize SQL execution efficiency.
5: Parallel Query improve query efficiency.
6: a detailed database table access statistics, data can distinguish between hot and cold, hot and cold data and can be separated automatically, so that the database is always in the best optimal thermal data read-write state.
7: columnar storage and data storage line storage exist, similar to excel, be able to calculate anyway, can read the database to achieve real HTAP, the perfect support OLTP and OLAP scene.
8: cluster management more streamlined and achieve fool-style management, namely the completion of the deployment of mature clusters, failover management more secure and reliable.
Three: hybrid database
The current database can be said that you have me, I have you, what you have, your support, I must also have, but also support. Various database has its own optimum application scenarios, but the scene is diverse, there have been Oracle, MySQL, PG, Redis, MongoDB, Flink, Kafak other databases exist, so that the development and operation and maintenance personnel are huge challenge, the more the use of more complex, just install the software you need to spend a lot of energy.
It may occur in the future A database that combines a variety of database expertise, and can modify the code zero set multiple databases for direct use, like MariaDB support Oralce databases, simply modify sql_mode = Oracle, can be fully compatible oralce syntax. Such a database could be a framework for other databases directly plug and play. The final one kind of database to dominate the world.
The idea is constantly updated. . .