Shulou(Shulou.com)06/01 Report--

If data is the "oil" of science and technology, then the database is the oil well. No matter which program or platform you use, you must have a database. The database is so important that all the giant enterprises are grabbing the database market. In the field of open source technology, databases have grown so fast that we can see a new database coming in almost every week. In this case, have you ever thought about which database will dominate the industry?

Analysis on the Development trend of Database

Let's first take a look at the development of the database. From the perspective of the whole database environment, specialization is the development direction of most enterprises. Our database should be able to solve specific problems, such as providing raw data query according to time series, and a real-time search scheme based on Lucene. But for some startups, their system architecture may not apply to this model. As a result, the native multi-model database is generated. The multi-model database represented by ArangoDB and Cosmos DB has been popular with many small and medium-sized enterprises.

Multi-model database has both key/value key / value pairs, graph graph and document document data model, provides a unified database query language covering three data models, and allows mixed use of the three models in a single query. This database is suitable for many different use cases, minimizes background components, supports different data modeling technologies (such as documents, charts, etc.), and helps enterprises reduce total cost of ownership and increase flexibility. in order to meet the overall technology stack requirements.

On the surface, multi-model database is almost the most perfect database, but in fact, this kind of database also has drawbacks, the database structure system is more closed, can not be personalized upgrade for specific use cases, the versatility is very poor. Due to the different types of workloads, we need a professional database. For example, most use cases of the Internet of things are writer-intensive. It is necessary to store data in chronological order and improve the performance of queries because it must have low latency when writing and being able to separate when reading.

So, in the end, how to choose a database, who will dominate the entire database market, in fact, it is difficult to find the standard answer. But if we have to build a database, then relatively speaking, a time series database will be the best choice, because in the long run, using a multi-model database will eventually lead to data migration problems.

About key-value database

So why do we need a key-value database? This may be an issue that everyone will be concerned about.

Most databases are basically modeled based on key-value databases. Because key value storage is more flexible. Neither a relational database like MySQL nor a graphical database like Dgraph has this capability.

Key-value database is a kind of non-relational database, which stores data by simple key-value method. The key-value database stores data as a collection of key-value pairs, where the key is the unique identifier. Keys and values can be anything from simple objects to complex composite objects. Key-value databases are highly partitioned and allow horizontal scaling at a scale that cannot be achieved by other types of databases. For example, a user can easily create a document database like MongoDB on a key-value database, where each field of the document will be mapped to a unique key.

The key-value database meets the modeling needs of the enterprise. Because it is only set up and operated at the bottom, performance tuning can be maximized. Therefore, the key-value database has become one of the best scalable databases, and one database can solve almost all specialized database needs.

But the question is, why didn't the key-value database finally dominate the world? That's because, compared with distributed databases and transactional databases, key-value databases have a lot of functions that they don't have.

FoundationDB has become a rising star.

FoundationDB, developed in 2009, is a distributed database that can store large-scale structured data on multi-cluster servers. The database system focuses on high performance, high scalability, and good fault tolerance. At the same time, FoundationDB is also a database of transactional key values arranged in lexical order. It is fully compatible with ACID, which means that our database (data and indexes) will always be in a consistent state.

On April 20, 2018, Apple announced that it will open source the core of its database product FoundationDB, which means that more and more enterprises will be able to implement higher-level database modeling in their own way. However, not everyone has the time or willingness to model the database, and that's what makes FoundationDB revolutionary.

FoundationDB's approach is to separate the data model from storage. For example, the data store does not have a built-in index. The upper layer provides the corresponding functionality, which implements the index by creating and storing two key-value pairs, one for data and one for indexing.

Summary

FoundationDB takes database competition to a new level. I can't say it's a perfect database because its underlying architecture is unknown. But FoundationDB does give more users an extra layer of choice, making it easier to operate on the basis of both flexibility and higher performance.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.