MySQL plus transaction processing equals MaxSQL_PHP tutorial

Last August at the OReilly-sponsored open source conference, MySQL developer Monty Widenius announced a new project called MaxSQL, an enhanced version of the currently popular MySQL database. The most important thing is that MaxSQL incorporates the latest Berkeley DB library from Sleepycat software, so the program supports transaction processing with another table type.

Currently, you cannot directly install the binary version of MaxSQL. You must compile it from the source code of MySQL 3.23 beta, but by the time you read this article, you should be able to download MaxSQL from the MySQL website. If you want to compile the latest version yourself, please see the article "Compiling your own MaxSQL", where I describe the compilation steps in detail.

What is a transaction?

Why does adding a transaction-safe table require a new project name? Perhaps I should mention that Monty Widenius, the main developer of MySQL, has two children, one named My and the other Max. Not only do the two brothers compete with each other, there is also a rival PostgreSQL.

About PostgreSQL, our other article in this issue "PostgreSQL adds a level to e-commerce" will mention that it is a competitor that is beneficial to us all, because every project team is overcoming Weaknesses in your own product. What MaxSQL does is bridge the gap between these two rivals. MySQL has fast read operations, so it is very popular with Web developers because Web content is mainly read. However, on the other hand, because MySQL uses ISAM tables, the entire table must be locked during write operations, which slows down update and insert operations, which is a big problem when the traffic is heavy. When the number of requests per second increases dramatically, write operations are queued in the server, and timeout errors occur on the Web page.

After using MaxSQL, you can still create ordinary fast ISAM type tables, or you can choose to use the new BDB type when a certain table needs to use transaction security features.

BDB tables are much faster than ISAM during write operations. Because they don't use table-level locks. In addition, the processing of any transaction table is logged in case of hardware failure. The log allows COMMIT/ROLLBACK operations. For example, if you place an order in an online store, the order will be recorded as a row in the "order" table. The system will subtract the corresponding inventory quantity from a row in the "inventory" table. If using MySQL's ISAM table, a CGI program needs to perform the following six steps:

LOCK inventory table
LOCK order table
UPDATE inventory table
UPDATE order table
UNLOCK inventory table
UNLOCK order table

If someone locked it For any table, this CGI program must wait. Once both are locked, CGI can update them and then release the lock. If step three fails (for example, the server is down), the order table will not be updated and your inventory will have been reduced.

After using the MaxSQL BDB transaction security type table, only four steps are needed:

BEGIN
UPDATE inventory table
UPDATE order table
COMMIT

You don't have to wait for the lock to be released, all four steps are one transaction. As soon as the BEGIN statement is read, MySQL reads the command into the buffer until it sees the COMMIT command. Therefore, all operations occur at the same time. Even if an unexpected operation occurs (disk full or power outage), the database will not be destroyed. In a non-transaction-safe system, if the third step fails, the database will be inconsistent. In the BDB table, if the operation on the order table fails, MySQL will resume the operation on the inventory table, so that no inconsistency will occur. .

Many websites use early versions of MySQL to implement table locks, but after using MaxSQL, it will be easy and fast.

What other new weapons are there?

In addition to the need to convert MySQL to MaxSQL because of the BDB table, there are also some important changes. One of the secret weapons is database replication. You can use one server as the master server and configure any number of slave servers, so that updates to the master server will also be copied to the slave servers. In order to use such a system, change your CGI scripts to make them aware of the existence of the MySQL slave server. This script will also switch to the slave server when it cannot connect to the master server.

It should be noted that such a system is only effective for databases with few write operations, because replication is not synchronous. If someone starts a transaction on the master server and the server dies before the transaction ends, the master server will not have time to copy the updates to the slave servers. In this way, this system is not suitable for e-commerce. Moreover, the CGI script always connects to the main server first, so it is not a load-balanced system.

Another major improvement is the format of ISAM type tables. The default table type is now called MyISAM. This type of table can handle up to 2GB per table and is cross-platform. You can copy a MyISAM file from Linux to Solaris without any conversion.

Conclusion:

MaxSQL can be used as an excellent alternative to expensive business systems. Many systems do not need to process many transactions and require Oracle to be installed. The transaction-safe tables provided by MaxSQL make the programming work of updating the database easier.

http://www.bkjia.com/PHPjc/532068.htmlwww.bkjia.comtruehttp: //www.bkjia.com/PHPjc/532068.htmlTechArticleAt the open source conference sponsored by OReilly last August, MySQL developer Monty Widenius announced a new project It's called MaxSQL, an enhancement of the currently very popular MySQL database...