Database Shadowing and Mirroring: How It Works

What Is Database Shadowing?

Database shadowing means that a database is able to work with two or more copies of itself at the same time. Among these copies, the original — referred to as the "master" — is the normal file from which data is retrieved. The other copies are called "shadows" or, at times, "mirrors." Whenever an alteration or change of any kind must be made to the master, the same changes must also be made to the shadows, so that each shadow is always up-to-date with the master copy. All changes must be made separately to each shadow so that there is no corruption of the master, and so that any physical corruption of the master is not inadvertently transferred to the shadows. (Resilience)

In the application Mimer SQL, shadowing is generally controlled at the databank level. In any databank where there is a minimum of transaction control, the database can be shadowed to as many copies as required at any given time. This means that important data can be well protected against corruption or disk crashes without having to rely on the traditional methods of maintaining backups. If a disk crash or failure occurs in the drive where the master data is stored, operations automatically continue using the shadows, without any interruption.

The key benefit of maintaining shadows is that work can continue uninterrupted. Without shadows, a disk failure would require a time-consuming process of restoring the master database from backup files stored on tapes or other media. With shadowing, the application is kept up and running at all times. The database manager can choose a convenient time, transform the shadows into master databanks, and create new shadows based on the updated data. Importantly, maintaining shadows has no impact on running time or response times. Background threads carry out the various changes made to the master databank and to the shadows without affecting the users of the application. (Resilience)

How Database Shadowing Works

A database shadowing system is able to perform both data check-in and check-out functions, and this is generally accomplished in several steps. First, the data stored in a database that must be edited is identified, so that copies or shadows can be made. Second, the relevant data is copied from the database to an active memory file, so that an active memory copy can be successfully isolated from the original data item. Third, a particular tag is associated with the active memory copy of the data item, or the state of that data item is changed.

The next step involves tagging, depending on the completed operation performed on the active memory copy of the data item. Finally, all changes made to the active memory copy are saved to the database data item by issuing a "Save" command. After this, the data item tag of the active memory copy controls the way in which the database items may be changed or transformed. (Database Shadowing System with data tags that select an operation of the save command)

Key Vendors: Mimer SQL and Others

There are several vendors of database shadowing systems, and Mimer is one of them. Mimer is a relational database management system developed by Sysdeco Mimer AB in Uppsala, Sweden. It is a high-speed database engine targeted at mission-critical production environments. Important features include scalable presentation and multi-processor support, along with availability on UNIX, Windows NT, Windows 95, Windows 98, and Open VMS platforms. Mimer is well suited to open environments in which interoperability is essential. Its key characteristics include high performance, ease of use, self-tuning capabilities, and built-in stability.

These features make Mimer one of the better choices for software products with an embedded DBMS intended for Internet, intranet, and extranet applications, where maximum database uptime and minimal database administration are real requirements. The program conforms 100% to general SQL standards and is also a native implementation of ODBC (Open Database Connectivity), the Microsoft interface. Mimer can be accessed by numerous development tools and is well suited to mission-critical multi-tier solutions. Its conformance to SQL standards makes it compatible with various transaction processing middleware products based on CORBA and OTM techniques. (Mimer 8, Product Overview)

Another company offering database shadowing systems is Imaging Solutions. Founded in 1994, the company states that its mission is to provide a complete suite of products and services needed to create an appropriate ECM (Enterprise Content Management) solution tailored to specific requirements in the area of database shadowing and mirroring. (Welcome, a Cautionary Tale)

Conclusion

Some people opine that shadowing alone is not an adequate solution, and that if a regional disaster or outage were to occur, the entire production environment may be affected. The Remote Data Recovery Facility used by many companies ensures that relevant data is not only well protected but also insulated against further risk of loss through the process of shadowing. This approach is an excellent solution for read-only functions such as data warehousing and mining, batch reporting cycles, complex SQL queries, load balancing, and local online access at the shadow site. Information and data can be safeguarded using the system of shadows for database applications. (Data Recovery without data loss)