Compute Cluster Load Balancing via Disk I/O Cache Sharing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In information handling systems, load balancing between physical nodes in a compute cluster is inefficient due to redundant storage and utilization of disk blocks, leading to suboptimal resource management and increased latency in virtual computing environments.
Innovation Solution
Implementing a method that caches frequently used disk blocks in memory and compares their contents across nodes, allowing for the elimination of redundant storage by redirecting virtual machines to hosts with existing copies, thereby optimizing resource allocation and increasing execution capacity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machines are distributed across physical nodes in a compute cluster, then resource utilization is improved, but redundant storage of disk blocks occurs leading to decreased efficiency
Solution Approach 1:
The patent merges disk block storage across multiple physical nodes by implementing a sharing mechanism. When a disk block is cached in memory on one node, other nodes can access and share that same cached block rather than maintaining separate redundant copies. This combining of storage resources eliminates redundancy while maintaining the distributed virtual machine architecture.
Solution Approach 2:
The patent makes disk block storage universal across the compute cluster by implementing a shared caching system. A single cached disk block in memory can serve multiple virtual machines across different physical nodes, making the storage resource multi-functional and eliminating the need for node-specific redundant copies.
2Speed
If disk blocks are cached in memory on each physical node, then I/O operation speed is improved, but redundant memory usage increases
Solution Approach 1:
The patent combines memory resources across physical nodes to create a shared disk block cache. Instead of each node maintaining separate cached copies of the same disk blocks, the system merges these caching functions so that one cached copy can be shared across multiple nodes, reducing total memory consumption while maintaining fast I/O access.
Solution Approach 2:
The patent implements a virtual copying mechanism where disk block references are shared across nodes rather than physically duplicating the actual block data in memory on each node. This allows rapid access to cached blocks across the cluster without proportionally increasing total memory usage.
3Device complexity
If load balancing is implemented without considering disk I/O cache contents, then system complexity is reduced, but resource management efficiency decreases
Solution Approach 1:
The patent implements feedback into the load balancing system by monitoring disk block caching status across physical nodes. The load balancing decisions are informed by real-time or near-real-time information about which disk blocks are cached where, allowing the system to optimize virtual machine placement based on actual cache contents rather than using simple static algorithms.
Solution Approach 2:
The patent performs preliminary actions by pre-fetching and caching disk blocks in memory before they are needed for I/O operations. The system anticipates which disk blocks will be needed based on virtual machine workloads and caches them in advance on appropriate physical nodes, improving I/O performance without waiting for actual access requests.
Data Source
AI summary
Load balancing reduces memory consumption in a virtual computing environment. Disk blocks in a cluster of hosts are monitored for redundancy. Execution of virtual machines may be migrated to different hosts within the cluster to balance loading.