SSD PCIe Link-State Control Across Four Performance Zones
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing SSD power management strategies, such as PCIe link speed modulation and PCIe link power state management, incur latency and transitional energy costs, and do not allow the backend to enter an idle state, resulting in suboptimal performance and quality of service.
Innovation Solution
A hybrid approach combining PCIe link speed modulation and PCIe link power state strategies to balance energy costs and quality of service, configuring four distinctive zones that provide high QoS at high performance points with graceful degradation towards lower performance points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If PCIe link speed modulation is used, then power consumption is reduced, but latency increases and quality of service deteriorates
Solution Approach 1:
The patent implements dynamic switching between link speed modulation and power state modulation based on real-time workload conditions. The system dynamically selects the optimal power management mode: using link speed modulation for sustained workloads to maintain low latency, and power state modulation for idle periods to maximize power savings, thereby resolving the contradiction between power consumption and latency
Solution Approach 2:
The system changes the operational parameters of the PCIe link by alternating between two different power management strategies. Link speed modulation changes the data transmission rate parameter, while power state modulation changes the link state parameter (L0/L1). This parameter switching allows the system to optimize for either power consumption or latency depending on current conditions
2Use of energy by moving object
If PCIe link power state modulation is used, then backend can enter idle state, but latency increases and quality of service deteriorates
Solution Approach 1:
The system dynamically adjusts the use of power state modulation based on workload characteristics. For sustained or unpredictable workloads, it prefers link speed modulation to avoid latency. For predictable idle periods, it uses power state modulation to achieve deeper power savings. This dynamic adaptation resolves the contradiction between enabling idle states and maintaining low latency
Solution Approach 2:
The patent employs periodic duty cycling where the link alternates between active and idle states in predetermined intervals. This periodic action allows the backend to enter idle states for power savings while maintaining the capability to quickly resume activity, balancing power consumption with latency requirements through rhythmic on-off patterns
3Use of energy by moving object
If link speed is reduced, then power consumption decreases, but backend performance decreases
Solution Approach 1:
The system dynamically adjusts link speed based on actual workload demands rather than using a fixed reduced speed. During high-performance needs, it maintains higher link speeds; during low-utilization periods, it reduces speeds for power savings. This dynamic adjustment resolves the contradiction by matching performance output to actual productivity requirements
Data Source
AI summary
Instead of the Peripheral Component Interconnect (PCI) Express (PCIe) link speed modulation and the PCIe link power state being utilized separately, the PCIe link speed modulation and PCIe link power state strategies are used in combination. The combination balances between the PCIe front end energy cost, the backend energy cost, and the quality of service (QOS). The performance/power space of the solid state drive (SSD) is mapped to the PCIe link speed modulation and PCIe link power state strategies to configure four distinctive zones, which provide high QoS at high performance points with a graceful performance degradation towards the lower performance points. A power/thermal constrained system dictates a certain performance that the SSD will be able to satisfy optimally by using both methods at different performance points.


