Newsql is more efficient than hadoop

Today, we have completely abandoned the traditional relationship. Database And busy using emerging nosql Data Library may not be a reasonable choice. On the contrary, the improved SQL (Structured Query Language) system may adjust some technical details. Distributed Database at the nosql August 23 Conference held in San Jose, California in 2011 Company Michael Stonebraker, chief technology officer of voltdb, expressed his views. 7 C3 H8 d % B, E2 A: O! D
The company where Stonebraker is located provides a newsql-based database. Software . The new architecture he advocated is more load-tolerant than the database software provided by traditional vendors. Stonebraker is the chief designer of Ingres and ipvs databases. He is also co-founder of vertica (for column database companies). HP has acquired vertica in February. (V f-B) P5 J0 :? % _-{
Compared with the booming development of nosql, the SQL-based relational database system is indeed a little dead. However, this is a database vendor error, not an SQL error. + M: W * q "}. K3 {6 G1 x (M
Stonebraker points out that most commercial database software has been on the market for 30 years or longer Time . They are not designed around automation, data weighting, and transactional environments. At the same time, the new Function Not as good as you think. /A' Q5 q0 C3] & L
Stonebraker indicates that the lag of the database system is usually attributed to multiple factors. For example, to restore Logs The buffer pool maintained for the target database system, as well as the data fields that are locked and locked. In voltdb Test These behaviors consume 96% of the system's resources.

The popularity of many emerging nosql databases, such as mongndb and Cassandra. This makes up for the limitations of traditional database systems. Consultant dan mccreary says the disadvantages of relational databases are stimulatingDevelopmentCreate a nosql database. Relational databases are not very flexible. Their basic architecture design is still in the punch card era, which reflects the strict data modeling method. If an organization needs to add data in another column, they must change the architecture, which may be tricky. The Relational Tables (entity models) created during modeling do not always reflect data accurately in the real worldHowExist.
McCreary also pointed out that another problem with SQL databases is that they do not have good scalability. When data increases by more than oneServerWhen the limit is reached, data must be shared or split to multiple servers. It is a complicated process to span multiple servers. In addition, external links may cause problems. For example, if data is converged in multiple tables, performing operations across servers may cause some problems .! O $ T9 @ 3 K3 D6 m
Stonebraker believes that nosql databases can provide good scalability and flexibility, but they also have their own shortcomings. Because SQL is not used, nosql database systems do not have highly structured queries and other features. Other problems with nosql include the inability to provide acid (atomicity, consistency, isolation, and durability) operations. In addition, different nosql databases have their own query languages, which makes it difficult to regulateApplicationProgramInterface.

Stonebraker indicates that newsql provides some unique features of SQL and also has the scalability of nosql. Newsql has a new architecture design. It releases the buffer pool that consumes system resources in databases running the main memory. The voltdb system uses the innovative newsql architecture, which is 45 times faster than traditional relational databases in transaction execution. Voltdb has 39 scalable servers and can process 1.6 million transactions per second (300 CPU cores ). Hadoop with the same processing capability requires more servers. For example, to perform the same tasks, voltdb requires 20 nodes and hadoop requires 1000 nodes to execute. /I3 G4 J + S6 ['u6 J' S6 Q6 |
the founder of DoubleClick and one of MongoDB's founders, Dwight Merriman and Stonebraker, agree that SQL itself is not the root cause of scalability and low performance. But Dwight Merriman also said that in the years to come, maybe not everyone would like to use SQL analyze and query their data. Because for for programmers , SQL-based stored procedures are particularly difficult.
at last, McCreary agreed with Stonebraker that nosql does not have a unified Query Language, which slows down the development of nosql. However, it is recommended that you use a language other than SQL in the new unified query tool of the database system. Such as XQuery, an XML document query language.