Memory Device Ring Power Control Under a Shared Power Budget
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic systems face inefficiencies in managing power consumption of multiple memory devices, leading to overdesign or compromised performance due to wide power swings, especially in data centers with many SSDs, which results in high initial and recurring costs.
Innovation Solution
A method for dynamically managing power among multiple memory devices by setting a predefined power budget and using peer-to-peer communication of power data packets to adjust individual device power levels, ensuring total system power remains within the budget while allowing each device to operate in its full power range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If electronic systems are overdesigned to accommodate worst-case power consumption of all memory devices operating at maximum power levels, then power supply and thermal management designs become compatible with maximum loads, but initial cost and recurring cost increase significantly
Solution Approach 1:
The patent implements dynamic power management where the host device continuously monitors and adjusts the power allocation to individual memory devices based on real-time system conditions. Instead of designing for static worst-case scenarios, the system dynamically adapts power distribution, allowing memory devices to operate at maximum power only when needed while keeping total system power within budget limits. This resolves the contradiction by making the power supply design flexible rather than fixed, maintaining reliability for peak loads while reducing costs for typical operation.
Solution Approach 2:
The system changes the power consumption parameter of memory devices dynamically through host-controlled adjustments. The host device modifies operational parameters such as cache memory usage, read-ahead operations, and power management modes of individual memory devices based on current workload and thermal conditions. This allows the system to operate efficiently under normal conditions while maintaining the capability to handle peak power demands, thus reducing initial and recurring costs without sacrificing reliability.
2Use of energy by stationary object
If a moderate limit is set for total power of memory devices in an electronic system, then power consumption is controlled within budget, but power for each individual memory device is limited and maximum performance cannot be achieved
Solution Approach 1:
The patent applies local quality by allowing different memory devices to operate at different power levels simultaneously. Instead of uniformly limiting all devices to a moderate power level, the host device individually adjusts power allocation based on each device's current workload, thermal state, and performance requirements. This enables some devices to operate at maximum power when needed while others operate at lower levels, maintaining total system power within budget while allowing individual devices to achieve maximum performance when necessary.
Solution Approach 2:
The system dynamically reallocates power among memory devices based on real-time conditions. When one device requires maximum performance, the host can temporarily allocate more power to that device while reducing power to other devices with lower current demands. This dynamic power sharing resolves the contradiction by making total power consumption controllable while still allowing individual devices to reach maximum performance when needed, rather than imposing static uniform limits.
3Use of energy by stationary object
If power levels of memory devices are dynamically adjusted based on system power level feedback, then system power level is maintained within predefined budget, but complexity of power management increases
Solution Approach 1:
The patent implements feedback-based power management where the host device monitors the total system power level and uses this information to adjust power allocation to individual memory devices. The system continuously receives feedback on power consumption and dynamically modifies operational parameters such as cache usage and read-ahead operations to maintain total power within the predefined budget. This feedback mechanism resolves the contradiction by automating power management, reducing the need for complex manual intervention while maintaining strict power budget control.
Data Source
AI summary
This application is directed to dynamic power management among multiple memory devices of an electronic system. A plurality of memory devices are coupled into a ring of memory devices, and passes a power data packet along a power control path that tracks the ring of memory devices continuously. During a current cycle, a first memory device receives the power data packet from an upstream memory device on the power control path, and the power data packet includes at least a system power level indicating total power consumption of the plurality of memory devices. The first memory device sets a current power level of the first memory device based on the received power data packet, updates the power data packet based on the current power level, and sends the updated power data packet to a downstream memory device on the power control path.