Router Parking for Power-Efficient Mesh Interconnects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In computer systems with multiple processing nodes, the interconnect fabric consumes a significant portion of power due to active routers maintaining packet forwarding capabilities even when nodes are in low power states, leading to high interconnect power consumption with low utilization.
Innovation Solution
Implementing a Router-Parking approach where selected interconnect routers are power-gated when associated nodes are in deep sleep states, using a centralized Fabric Manager to proactively aggregate traffic and maintain network connectivity while minimizing latency and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If routers remain always active to maintain packet forwarding capability, then network reliability is improved, but power consumption increases
Solution Approach 1:
The router transitions between two operational states: active and parked. When a node enters deep sleep state, its associated router changes its operational parameters by entering a parked state with minimal power consumption, yet maintains the capability to be reactivated. This state transition allows the system to reduce power consumption dynamically while preserving network reliability when needed.
Solution Approach 2:
The router parking mechanism introduces dynamic adaptability to the interconnect fabric. Routers dynamically adjust their operational state based on the activity state of associated nodes. When nodes are inactive, routers park themselves; when nodes become active, routers reactivate. This dynamic behavior allows the system to optimize power consumption in real-time while maintaining network reliability when required.
2Use of energy by moving object
If more nodes are placed in low power states, then power consumption is reduced, but network connectivity may be degraded
Solution Approach 1:
The fabric manager acts as an intermediary that coordinates router parking decisions across the network. It monitors node states and orchestrates which routers should park and which should remain active to maintain network connectivity. This centralized coordination ensures that even when many nodes are in low power states, the network maintains adequate connectivity through strategically kept-active routers.
Solution Approach 2:
The system dynamically adjusts the operational parameters of routers based on overall network load and connectivity requirements. When many nodes are inactive, more routers can park; when connectivity requirements increase, routers reactivate. This parameter adjustment allows the system to maximize power savings while maintaining acceptable network connectivity levels.
3Loss of energy
If routers are parked to reduce power consumption, then static energy is reduced, but latency may increase
Solution Approach 1:
Routers perform preliminary actions by maintaining their basic forwarding capability in a parked state without fully shutting down. They remain in a low-power state but retain the ability to quickly reactivate when needed. This preliminary preparation allows them to reduce static energy consumption while minimizing latency impact when traffic needs to be forwarded.
Solution Approach 2:
The router parking mechanism operates periodically based on node activity patterns. Routers monitor their associated nodes and periodically adjust their state accordingly. This periodic operation allows the system to capture power savings during extended idle periods while minimizing latency impacts during active periods when nodes wake up and require network communication.
Data Source
AI summary
A method and apparatus for selectively parking routers used for routing traffic in mesh interconnects. Various router parking (RP) algorithms are disclosed, including an aggressive RP algorithm where a minimum number of routers are kept active to ensure adequate network connectivity between active nodes and/or intercommunicating nodes, leading to a maximum reduction in static power consumption, and a conservative RP algorithm that favors network latency considerations over static power consumption while also reducing power. An adaptive RP algorithm is also disclosed that implements aspects of the aggressive and conservative RP algorithms to balance power consumption and latency considerations in response to ongoing node utilization and associated traffic. The techniques may be implemented in internal network structures, such as for single chip computers, as well as external network structures, such as computing clusters and massively parallel computer architectures. Performance modeling has demonstrated substantial power reduction may be obtained using the router parking techniques while maintaining Quality of Service performance objectives.