CC Search Sitemap
High Availability  / Components  / Application Availability  / Cluster  / High Availability Cluster  / SteelEye LifeKeeper  / Advanced Appliance Scenarios  /  Continuous Data Protection (CDP)

Continuous Data Protection

Continuous Data Protection (CDP) is the continuous backup of data which have been modified and provided with a time stamp or another label. Following logical errors such as a destruction through viruses, a corruption through software faults or a deletion by the user, CDP will enable the affected data to be restored at a suitable point before being exposed to the error. This allows the user to develop and test a temporary data record and, if necessary, to search backwards and forwards until a consistent data stock has been established.

CDP Basics

Prinzip Continuous Data Protection
Click to enlarge image

Continuous Data Protection differs from conventional data protection in two main ways:

  • Each change of a primary data stock will also lead to a change in the Backup data stock. This ensures that a copy of the data stock is always available in real-time.
  • The data flow changing the data is recorded using the original data stock. Significant places in the data stream can be marked using a Marker.

Because it contains additional information, a greater amount of data is used for the CPD than is contained in the primary data stock. The original data stock plus the data flow up to a specific Marker equals the data stock which corresponds with the Marker at that time.

Markers can be set through application software (e.g. when a database is being reorganised or maintained, or when a data record is being written) or manually by the operator (e.g. when Patches are being installed). These Markers can also be used to navigate in the data stream.

When the available storage space has been filled by the CDP data stock, the original data are merged with the oldest recorded data to create a new basic data stock. The thus emptied storage space is then made available for storing new data records. The oldest data stock which is restored in this manner is the newly generated basic data.

Using CDP

Continuous Data Protection (CDP) is an addition and extension of conventional Backup. Whilst Backup secures data intermittently, CDP continuously updates the data stock. Markers are set in the data stream at specific times or during specific events allowing these to be easily found again later. Markers also designate consistent data stock.

The CDP continuous data protection method can be used to repair damage arising from technical faults such as equipment breakdowns, faulty Patches or Updates as well as from logical errors such as the deletion of databases and data records or directories. Here it is important to use as many expressive Markers as possible. In the event of the destruction of data, all work on original data must be interrupted.

At first only the time when the fault was discovered is known. The time of the damage is determined by establishing the time when the data were most likely still correct. The objective is to find the time when the data were still correct and consistent directly before the damage occurred. For this purpose the data stock is combed by producing a test data stock up to a specific Marker using a copy of the basic data stock and the data flow. This test data stock is then tested with the application and evaluated.

After the best possible data stock has been established in this manner, it is then copied on the source side after which work can continue with a minimum of data loss. This data stock then constitutes the original data for the further CDP function.

CDP-Navigation

Navigation im Datenbestand - Continuous Data Protection
Click to enlarge image

Best possible result of restored data stock using the CDP

Bestmögliche Daten - Continuous Data Protection
Click to enlarge image

The chart shows the best possible result of restored data stock using the CDP method together with the time and function Markers.

Proceeding from the damage period which is initially defined by establishing the fault and the last known correct data stock, the point before the data was destroyed must be established by purposeful testing so that as few data as possible are lost.