Virtual Machine Ballooning for Grid Latency Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Performance issues in virtual machines hinder their adoption in grid computing environments due to high latency between request and availability, making them unattractive for complex or numerous computing tasks.
Innovation Solution
Pre-creating virtual machines in a tiny configuration, allowing initialization before shrinking, and ballooning to full size upon request, while maintaining a pool of live tiny virtual machines to reduce resource consumption and latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If virtual machines are pre-created in full size, then requests can be granted immediately, but resources are consumed continuously even when not in use
Solution Approach 1:
The virtual machine transitions dynamically between two states: a tiny pre-created state for immediate availability and a full expanded state when activated. This dynamic transformation allows the system to optimize between readiness and resource efficiency, consuming minimal resources during idle periods while providing full functionality when needed.
Solution Approach 2:
The virtual machine is pre-created in a minimal configuration before actual use, allowing the initialization and basic setup to be completed in advance. This preliminary action reduces the activation time when a request is made, while the full configuration is only established when truly needed, avoiding continuous resource consumption.
2Use of energy by moving object
If virtual machines are created on-demand, then resources are consumed only when needed, but there is significant initialization delay when requests arrive
Solution Approach 1:
A minimal virtual machine instance is created and initialized in advance, maintaining only the essential core components in a ready state. This preliminary action ensures that when a request arrives, the activation delay is minimized since the basic infrastructure is already in place, while full resource allocation occurs only when needed.
Solution Approach 2:
The system transitions from a static on-demand creation model to a dynamic hybrid model where a minimal instance exists continuously and expands to full capacity upon activation. This dynamic approach balances resource efficiency with response time, maintaining a small footprint during idle periods while providing rapid activation when required.
3Productivity
If virtual machines are maintained at full size, then performance is optimal, but resource overhead is high during idle periods
Solution Approach 1:
The virtual machine size is dynamic rather than static, adjusting between a minimal configuration during idle periods and a full configuration during active use. This ensures optimal performance when the virtual machine is running while minimizing resource overhead when it is not in use, eliminating the need to maintain full-size resources continuously.
Solution Approach 2:
The system changes key parameters of the virtual machine configuration based on its operational state. During idle periods, parameters such as memory allocation, CPU resources, and storage are reduced to minimal levels. When activated, these parameters are expanded to full capacity, ensuring optimal performance without maintaining high overhead during idle times.
4Use of energy by moving object
If virtual machines are shrunk to tiny configuration, then resource consumption is minimized, but the machine may not have sufficient resources for immediate task execution
Solution Approach 1:
The virtual machine configuration is dynamic, transitioning from a tiny state optimized for resource efficiency to a full state optimized for capability. The tiny configuration maintains only the essential kernel and initialization structures, while the full configuration provides complete hardware virtualization and system resources. This dynamic adaptation ensures the machine has sufficient capability when needed while consuming minimal resources during idle periods.
Solution Approach 2:
The essential core components and initialization structures are pre-configured in the tiny virtual machine, ensuring that the fundamental adaptability and functionality are already in place. When activated, the remaining resources and configurations are rapidly allocated, providing full machine capability without requiring complete pre-configuration, thus balancing resource consumption with adaptability.
Data Source
AI summary
Virtual machines can be pre-created in a grid environment. Pre-created machines can be of a tiny configuration to avoid consuming resources. A request for a virtual machine in the grid can be granted by ballooning a pre-created tiny virtual machine to a larger configuration. The tiny pre-created virtual machine can be advertised as being of a non-tiny configuration. Agents can work in concert to coordinate handing requests for a virtual machine in the grid. The virtual machine can be initialized before it is reduced to a tiny configuration. Ballooning the virtual machine can take considerably less time than creating it from scratch. Thus, high performance computing can be provided on virtual machines in a grid environment, enabling virtual machine technology to be applied in a wide variety of grid-based scenarios.