Mass Storage Subsystem Data Layout for Mixed Workloads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems are optimized for either read or write operations, leading to suboptimal performance for mixed workloads, and lack the ability to provide differentiated service levels for data objects with varying service level requirements (SLRs).
Innovation Solution
A data layout system that divides storage into two divisions based on physical properties of storage devices, prioritizing read performance for high SLR data objects by using primary and spillover data chunks, while maintaining write anywhere capability and supporting multiple SLRs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data is stored in a write-anywhere file system to achieve high write efficiency, then write operations are optimized, but read operations become slower due to data fragmentation
Solution Approach 1:
The patent segments the mass storage subsystem into two distinct divisions: a first division optimized for read operations and a second division for write operations. This segmentation allows each division to be optimized for its specific function, resolving the contradiction between write efficiency and read speed by preventing read-degrading fragmentation in the first division while maintaining write flexibility in the second division.
Solution Approach 2:
The patent applies local quality by creating different data layout characteristics in different parts of the storage system. The first division uses a read-optimized layout with contiguous or near-contiguous data chunks, while the second division maintains write-anywhere flexibility. This local differentiation allows each region to excel at its intended operation type.
2Productivity
If existing storage systems are optimized for a particular workload type, then performance for that workload is improved, but performance for other workload types deteriorates
Solution Approach 1:
The patent creates a multi-functional storage system where the first division provides read-optimized performance and the second division provides write-anywhere flexibility. Together, these divisions create a universal storage system that can efficiently handle both read-intensive and write-intensive workloads, as well as mixed workloads, without sacrificing adaptability to different workload types.
Solution Approach 2:
By segmenting the storage system into two functional divisions, the patent enables the system to adapt to different workload types by directing read operations to the first division and write operations to the second division, thereby maintaining both specialization and versatility.
3Adaptability or versatility
If data chunks are separated by large distances to allow write anywhere operation, then write flexibility is maintained, but read performance deteriorates due to increased seek time
Solution Approach 1:
The patent segments the storage space into read-optimized chunks in the first division and write-flexible chunks in the second division. The read-optimized chunks are positioned to minimize seek distances, while the write-flexible chunks can be placed anywhere available. This segmentation resolves the contradiction by assigning different spatial characteristics to different functional requirements.
Data Source
AI summary
A system and method for allocating data objects across multiple physical storage devices in a mass storage subsystem first determines a set of physical properties associated with the physical storage devices. The system assigns portions of the mass storage subsystem to a first division or a second division based on the physical properties, such that read operations directed to logically related data stored in the first division can be executed more efficiently than read operations directed to data stored in the second division. During operation, the system stores data objects with a low SLR in the second division, which may be allocated according to any well-known file system. If a write request is for a new data object with a high SLR, the system stores the new data object in a set of neighboring primary data chunks in the first division. For subsequent write requests that modify a stored data object, the system stores the modified data in a spillover data chunk located in proximity to the primary data chunk.