Method of improving performance of a data storage device
View Patent ↗Methods are provided for efficiently storing data to a data storage device or subsystem. The data storage device may be a Solid-State Device (SSD), and may be implemented as part of a RAID (Redundant Array of Independent Disks) or other subsystem. When existing data is read and updated, and must be re-stored, the data is assembled and stored as if it were new data, and is written in a sequential manner, instead of being written to the same storage location. A newer generation number distinguishes it from the previous version. If the storage subsystem employs data striping, stripe size may be matched with the size of a logical collection of data (e.g., an extent), so that each such logical collection of data is wholly stored on just device in the storage subsystem. Concurrent device access may be supported by concurrently writing substripes of data to each device/extent.
1. A method of storing data, the method comprising:
executing a data reduction application to reduce data exchanged between a client and a server engaged in a client-server communication application;
receiving updated data for the data reduction application at a data storage system having a minimum physical write size equal to multiple logical pages of data and characterized by a threshold number of updates to stored data;
reading a logical page of data from a first physical offset within the data storage system, wherein the logical page of data includes data corresponding to the updated data;
constructing in memory a first new logical page of data to include:
the updated data;
new data not already stored in the data storage system; and
less than all content of the logical page of data;
assigning a new version identifier to the first new logical page;
buffering the first new logical page until a threshold number of new logical pages of data, including the first new logical page, are accumulated; and
once the threshold number of new logical pages of data have been accumulated,
sequentially writing the first new logical page to a current physical offset within the data storage system instead of to the first physical offset,
calculating extent metadata for the threshold number of new logical pages, and
writing the extent metadata to the end or to the beginning of the threshold number of new logical pages.
2. The method of claim 1 , wherein said threshold number of new logical pages of data comprise an amount of data matching the minimum write size of the data storage system.
3. The method of claim 1 , wherein the data storage system comprises one or more solid-state data storage devices.
4. The method of claim 1 , wherein said constructing comprises updating the logical page of data.
5. The method of claim 4 , wherein said updating comprises updating metadata within the logical page of data.
6. The method of claim 4 , wherein said updating comprises adding data to the logical page of data.
7. The method of claim 1 , wherein said constructing comprises combining the content of the logical page with content of one or more other logical pages.
8. A non-transitory computer-readable medium storing instructions that, when executed by a computer, cause the computer to perform a method of storing data, the method comprising:
executing a data reduction application to reduce data exchanged between a client and a server engaged in a client-server communication application;
receiving updated data for the data reduction application at a data storage system having a minimum physical write size equal to multiple logical pages of data and characterized by a threshold number of updates to stored data;
reading a logical page of data from a first physical offset within the data storage system, wherein the logical page of data includes data corresponding to the updated data;
constructing in memory a first new logical page of data from to include:
the updated data;
new data not already stored in the data storage system; and
less than all content of the logical page of data;
assigning a new version identifier to the first new logical page;
buffering the first new logical page until a threshold number of new logical pages of data, including the first new logical page, are accumulated; and
once the threshold number of new logical pages of data have been accumulated,
sequentially writing the first new logical page to a current physical offset within the data storage system instead of to the first physical offset,
calculating extent metadata for the threshold number of new logical pages, and
writing the extent metadata to the end or to the beginning of the threshold number of new logical pages.
9. A method of sequentially storing randomly read data, the method comprising:
executing a data reduction scheme to reduce data exchanged between two nodes in a communication network;
reading a first page of data from a first offset within a data storage system;
updating the first page of data to form a second page of data, wherein said updating includes discarding obsolete data that are no longer used in the data reduction scheme;
calculating a current offset within the data storage system, wherein the current offset is an offset at which a new sequential write would be performed;
if the data storage system has a minimum write size greater than a size of the second page of data, accumulating pages of data, including the second page of data, until a total size of the accumulated pages of data match the minimum write size;
once the total size of the accumulated pages of data match the minimum write size,
writing the second page of data to the current offset,
calculating extent metadata for the accumulated pages of data, and
writing the extent metadata to the end or to the beginning of the accumulated pages of data.
10. The method of claim 9 , further comprising:
assigning to the second page of data a generation identifier newer than a generation identifier of the first page of data.
11. The method of claim 9 , wherein said updating comprises adding data to the first page of data.
12. The method of claim 9 , wherein said updating comprises modifying data within the first page of data.