Dynamic SSD Data Stitching for Log Priority and Metadata Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current solid-state array (SSA) configurations statically partition SSDs into log, cache, and metadata sets, leading to inefficient utilization and performance imbalance, as usage is bi-modal and not aligned with IOPS and bandwidth requirements, resulting in wasted performance and increased latency due to metadata storage on disks.
Innovation Solution
A method and system that prioritize I/O service for log data over cache and metadata, dynamically stitch data types across multiple SSDs in a virtual pool, configuring variable percentages for cache and metadata on top of a fixed percentage for log data, and create multiple copies of log and metadata data while maintaining a single copy of cache data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If SSDs are statically partitioned into log, cache, and metadata sets, then the system structure is simple and easy to manage, but the performance utilization is inefficient and imbalanced due to bi-modal usage patterns
Solution Approach 1:
The patent implements dynamic partitioning where the metadata and cache data percentages are adjusted based on actual workload demands rather than fixed static allocations. The system monitors usage patterns and automatically reconfigures the partition sizes to match current performance requirements, transforming the static partitioning structure into a dynamic one that adapts to changing workload conditions.
2Device complexity
If metadata is stored on disks, then the system structure is simple, but the access latency increases and overall I/O performance deteriorates
Solution Approach 1:
The patent extracts metadata storage from the general-purpose disk storage and creates a dedicated metadata storage area within the SSD architecture. This separation allows metadata to be stored and accessed independently from regular data operations, reducing the overhead and latency associated with metadata access during I/O operations while maintaining a relatively simple overall system structure.
3Quantity of substance
If more SSDs are added to increase storage capacity, then the storage capacity increases, but the system cost and power consumption increase proportionally
Solution Approach 1:
The patent implements a multi-functional SSD architecture where each SSD device can serve multiple purposes: storing log data, cache data, and metadata simultaneously through dynamic partitioning. This allows the system to maximize the utilization of each SSD's capacity and performance, reducing the need to add additional SSDs to increase storage capacity, thereby lowering both system cost and power consumption.
4Stability of the object's composition
If static partitioning is used with fixed percentages, then the configuration is stable and predictable, but it cannot adapt to varying workload demands and wastes available performance capacity
Solution Approach 1:
The patent implements dynamic partitioning where the metadata and cache data percentages are adjusted based on actual workload demands rather than fixed static allocations. The system monitors usage patterns and automatically reconfigures the partition sizes to match current performance requirements, transforming the static partitioning structure into a dynamic one that adapts to changing workload conditions.
Data Source
AI summary
A disclosed system and method executable on a computer for managing a memory storage system includes prioritizing log data over cache data and over metadata in a configuration of multiple SSD (solid state drive) devices stitched in a virtual pool of data types including log write data, cache read data and metadata. The method also includes stitching the data types in the configuration across the multiple SSD devices in the virtual pool configured for a variable percentage cache data and a variable percentage metadata on top of a fixed percentage log data. The method further includes protecting the stitched data type configuration of multiple SSD devices in the virtual pool by creating multiple copies of the log data and multiple copies of the metadata and a single copy of the cache data. An intelligent and dynamic adjustment of metadata and cache data is based on a growth of the metadata.