Content locality-based caching in a data storage system

Abstract

A data storage caching architecture supports using native local memory such as host-based RAM, and if available, Solid State Disk (SSD) memory for storing pre-cache delta-compression based delta, reference, and independent data by exploiting content locality, temporal locality, and spatial locality of data accesses to primary (e.g. disk-based) storage. The architecture makes excellent use of the physical properties of the different types of memory available (fast r / w RAM, low cost fast read SSD, etc) by applying algorithms to determine what types of data to store in each type of memory. Algorithms include similarity detection, delta compression, least popularly used cache management, conservative insertion and promotion cache replacement, and the like.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]This application is a continuation of U.S. patent application U.S. Ser. No. 13 / 366,846 entitled Pre-Cache Similarity-Based Delta Compression for use in a Data Storage System, filed Feb. 6, 2012, which is hereby incorporated herein by reference in its entirety.[0002]U.S. patent application U.S. Ser. No. 13 / 366,846 claims the benefit of the following provisional applications, each of which is hereby incorporated herein by reference in its entirety: U.S. Ser. No. 61 / 441,976 entitled Intelligently Coupled Array of SSD and HDD, filed Feb. 11, 2011; U.S. Ser. No. 61 / 447,208 entitled Effective Page Classification and Delta Encoding AND Buffer Cache—Temporal and Content Localities, filed Feb. 28, 2011; and U.S. Ser. No. 61 / 497,549 entitled Conservative Insertion and Promotion Cache Replacement Algorithm, filed Jun. 16, 2011.[0003]U.S. patent application Ser. No. 13 / 366,846 is a continuation-in-part of U.S. patent application Ser. No. 12 / 762,993 e...

AI Technical Summary

Benefits of technology

[0009]Recent developments of flash memory-based solid-state drives (SSDs) have been very promising, with rapid increases in capacity and decreases in cost. Because SSDs are based on semi-conductor technology, they may provide great advantages including enabling high-speed random reads, low power consumption, compact size, shock resistance, and the like. However, limitations that result from the physical properties of SSDs include write and erase times that are very slow compared to read times, and a significantly limited number of write / erase cycles before failure of a block.

[0010]The above physical properties and operational characteristics of SSDs present unique challenges in designing a SSD-based mass storage hierarchy. Therefore, we disclose herein cache management methods and systems that can address these unique challenges by effectively managing an SSD-based storage hierarchy to provide higher I / O performance, lower cost, longer durability, and higher data reliability.

[0011]Also, while the capacity of disk drives grows rapidly, their electromechanical parts have held down the improvement of their performance. Caching plays a critical role in modern systems, the cache bridging the gap between a disk drive and the main memory. Applying content locality techniques to cache design may provide a significant improvement in performance, particularly when combined with the techniques for SSD use and optimization described herein. Content locality refers to the characteristic that many data blocks in a data storage system may share similar or even the same content.

Problems solved by technology

However, limitations that result from the physical properties of SSDs include write and erase times that are very slow compared to read times, and a significantly limited number of write / erase cycles before failure of a block.

The above physical properties and operational characteristics of SSDs present unique challenges in designing a SSD-based mass storage hierarchy.

Also, while the capacity of disk drives grows rapidly, their electromechanical parts have held down the improvement of their performance.

This may result in much more data redundancy than would be present when running a single OS on a data center server.

Images