Power-Aware Application Placement in Virtualized Servers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current workload placement on servers is inefficient, leading to sub-optimal resource utilization and high operational costs due to conservative provisioning approaches that prioritize hardware isolation and security over performance, resulting in low server utilization rates (averaging 11-50%) and increased energy consumption.
Innovation Solution
The development of a power-aware application placement framework, pMapper, which dynamically places applications on virtualized servers to minimize power consumption and migration costs while meeting performance guarantees by using techniques like First Fit Decreasing bin packing and incremental FFD, and considering power and migration costs across time windows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conservative provisioning approach with hardware isolation is used, then security and reliability are improved, but resource utilization deteriorates
Solution Approach 1:
The patent merges multiple applications onto shared physical servers through virtualization technology, replacing the conservative hardware isolation approach. Virtual machines provide logical isolation while enabling physical resource sharing, thus improving resource utilization while maintaining security and reliability through virtualization-based isolation mechanisms.
Solution Approach 2:
The patent makes physical servers universal by enabling them to host multiple different applications simultaneously through virtualization. Instead of dedicating entire servers to single applications, the same physical infrastructure can dynamically serve multiple workloads, improving overall resource utilization while maintaining isolation through virtual machine boundaries.
2Reliability
If conservative provisioning approach is used, then security is improved, but operational costs worsen
Solution Approach 1:
By merging multiple applications onto shared physical servers through virtualization, the patent reduces the total number of physical servers required. This consolidation decreases power consumption, cooling requirements, and operational costs while maintaining security through virtualization-based isolation mechanisms that provide adequate security without requiring separate physical hardware for each application.
3Use of energy by stationary object
If dynamic application placement is implemented, then power consumption is reduced, but migration cost increases
Solution Approach 1:
The patent implements dynamic application placement that adapts to changing workload conditions and power states. Applications are placed on servers in dynamic sleep or idle states when possible, and migration occurs only when necessary to optimize power consumption. The system balances dynamic placement benefits against migration overhead by making placement decisions based on current system state and predicted workload patterns.
Solution Approach 2:
The patent performs preliminary placement planning to minimize future migrations. By anticipating workload patterns and server power states, the system pre-positions applications on servers that are likely to remain in low-power states, thereby reducing the frequency and cost of migrations while maintaining power optimization benefits.
4Productivity
If server consolidation is increased, then resource utilization is improved, but power consumption worsens
Solution Approach 1:
The patent implements dynamic placement strategies that respond to changing power states and workload conditions. Instead of static consolidation, the system continuously monitors server power consumption and workload demands, dynamically migrating applications to optimize the balance between resource utilization and power consumption. This allows the system to consolidate workloads when power costs are low while maintaining utilization benefits.
Data Source
AI summary
N applications are placed on M virtualized servers having power management capability. A time horizon is divided into a plurality of time windows, and, for each given one of the windows, a placement of the N applications is computed, taking into account power cost, migration cost, and performance benefit. The migration cost refers to cost to migrate from a first virtualized server to a second virtualized server for the given one of the windows. The N applications are placed onto the M virtualized servers, for each of the plurality of time windows, in accordance with the placement computed in the computing step for each of the windows.