Publication Date: 2014-Jun-11
The IP.com Prior Art Database
Restore are critical process for data management, how to speed up these process become more and more important with the data scale extending repidly, traditional method is using deduplication process or offline restore method, these mothod have some shortages with the speed and cost conflict, based on this situation, there is new deduplication method appears to make higher effective restore scheme.
Page 01 of 5
To speed up restore process, currently method is to use deduplication process or offline restore method . This method is to provide a cache mechanism to reduce duration of restore. Cache can be made of memory or high speed disk and reduce the amount of throughput of transfer from backend storage server.
Deduplication method can be used to reduce duplicated trunks in each restore , restore cache can be used to further reduce
workload between frequent restore. Filespaces always be restored each time can be stored in cache, only modified filepsaces need to be retrieved from storage server.
The core idea of this invention is create the restore cache in the storage client side when setting up client , the files contains changed data will be recorded and saved in the cache, if restore operation be triggered, it will compare the original data with the cached data, pick the changed data, then send its file name into a file list in the client side, the file list will be fed back to server side, and only restore the data contained in the file list. It could reduce the restoring workload and improve the restore efficiency at great extent. On the other hand, we also design other two major functions to enhance the efficiency of this innovation , The User Behavior Analysis and Cache Update. The User Behavior Analysis is a standard to decide what data need to be copied to the cache, it will use the customer's restore frequency as the standard. The Cache Update could make sure that files with the highest restore frequency always remain in the cache, and keep data synchronization between the storage server and the cache. Comparing with traditional deduplicated restore mode, restore cache could further reduce workload between frequent restore, and the cache could be made of memory or other high speed device, it could also speed up the data processing more than the traditional mode. Does it sounds like a fantastic solution, right?
Restore cache contains three concepts: Differential Restore, User Behavior Analysis and Cache Update.
Differential Restore is aimed to shorten the restore time from server to client. To achieve this, it just needs to restore the changed files from server side.
Page 02 of 5
Figure-1 Differential Restore The procedure of differential restore like bellow:
Step1: Set a cache on the client system (after the storage client was installed on customer's system).
Step2: Copy customer's changed data to the cache.
Step3: Before restoring, compare the original data with the cached data to get the changed files name into a filelist . Step4: Only restore the changed files through the filelist.
User Behavior Analysis
User Behavior Analysis is a standard to decide what data need to be copied to the cache. In this paper, it will use the customer's restore frequency as the standard. File's restore frequency...