Checkpoint processing with save-while-active

Checkpoint processing occurs after the server determines exactly which objects it will save for a particular library. If the save-while-active request is for multiple libraries, then the server performs checkpoint processing for all libraries in the save request.

Checkpoint processing does not require that the server maintain two complete copies of the objects you are saving. The server only maintains two copies of the pages of the object that the applications are changing while you are performing the save. The more pages that applications change for an object during the save-while-active request, the greater the storage requirement for the object. After the server completes checkpoint processing to create the checkpoint image of the page, performance decreases slightly for the first update to a page. The performance impact varies depending on the disk type, available disk storage, and processor model. Further updates to the same changed page do not require any additional processing with respect to the checkpoint version of the page.

The following figure shows how the server maintains a checkpoint image of an object during a save-while-active operation. The shaded parts of the diagram represent the checkpoint version of the object. An explanation of the steps follows the figure.
Figure 1. Server management of updates to objects after checkpoint processing is complete
Checkpoint processing illustration

The figure above shows a timeline with T1 — T5:

  1. Time T1 is the save preprocessing phase of the save-while-active operation. The object reaches a checkpoint at the end of time T1.
  2. Time T2 shows an update to the object, referred to as C1. The update occurs while the save-while-active request saves the object to the media.
    1. An application makes a request to update C1.
    2. The server first makes a copy of the original page.
    3. The applications make the change to the object.

    The original page copied is then part of the checkpoint image for the object.

  3. Time T3 shows that the object received two additional changes, C2 and C3. Any additional change requests that are made to the pages of the object already changed for C1, C2, or C3 do not require any additional processing. At the end of time T3, the save-while-active request has completely saved the object to the media.
  4. Time T4 shows that the server no longer maintains copied pages for the checkpoint image of the object because the server no longer needs them.
  5. Time T5 shows the object on the server has the C1, C2, and C3 changes. But the copy, or image, of the object saved to the media does not contain those changes.
Related concepts
Commitment control with save-while-active