Smart Weighted Container Data Cache Eviction for SSD Bottlenecks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data storage systems, the limited capacity and high cost of SSDs in PIP storage systems create bottlenecks, leading to performance degradation due to data cache eviction, especially in low IOPS environments, which hampers Instant Access/Instant Recovery use cases.
Innovation Solution
Implementing a smart weighted container data cache eviction policy that weights WEUs based on page statistics, such as hit frequency and recency, to preserve frequently and recently accessed data, evicting less frequently and recently used data blocks only when necessary, and optimizing SSD usage to maintain high cache hit rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If SSD capacity is increased to improve data cache performance and reduce bottlenecks, then cache hit rate and IOPS are improved, but cost increases and write-erase cycle limitations are reached faster
Solution Approach 1:
The patent changes the parameter of cache eviction from traditional LRU (least recently used) to a weighted scheme that considers both recency and frequency of access. This allows more efficient utilization of limited SSD capacity by preserving frequently accessed data blocks, thereby improving cache hit rate without requiring increased SSD capacity.
Solution Approach 2:
The patent implements feedback mechanisms by tracking access patterns (hit frequency and recency) of data blocks in the cache. This feedback information is used to dynamically adjust eviction decisions, allowing the system to adapt to changing workload patterns and optimize cache performance within the constrained SSD capacity.
2Loss of time
If data cache size is increased to support IA/IR use cases with low IOPS, then read latency is reduced, but cost increases and the system becomes less economical
Solution Approach 1:
The patent modifies the cache eviction policy parameters to weight both recency and frequency of access, creating a more selective eviction strategy. This allows the limited SSD cache capacity to be used more effectively for IA/IR workloads, reducing read latency without requiring additional SSD capacity.
3Device complexity
If traditional LRU eviction policy is used to simplify cache management, then device complexity is reduced, but cache performance deteriorates due to eviction of frequently accessed data
Solution Approach 1:
The patent extends the basic LRU eviction policy by adding a frequency weighting parameter. This modification maintains the relative simplicity of LRU while incorporating frequency information to prevent eviction of frequently accessed blocks, thereby improving cache hit rate without significantly increasing policy complexity.
Data Source
AI summary
Smart weighted container data cache eviction preserves write evict units (WEUs) containing the most frequently and recently accessed blocks to maintain low latency data cache. Prior to performing cache eviction, the WEUs are weighted based on the page statistics maintained for each WEU. Page statistics include page hit/frequency and recency statistics associated with each WEU and data cache eviction is performed at the WEU level of granularity. Therefore, an entire WEU can be evicted based on page hit/frequency and recency statistics associated with the WEU.