Multi-Resource Schedulable Unit for Cloud Data Centers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of dynamically allocating multiple types of computing resources in a cloud computing environment leads to immense computational overhead, as existing methods require constant reallocation to meet peak demands while minimizing idle resources during non-peak times, resulting in inefficient resource utilization.
Innovation Solution
The introduction of a Multi-Resource Schedulable Unit (MRSU) that dynamically determines a set of correlated computing resources (CPU cores, RAM, storage, and communication bandwidth) based on application needs, allowing for coarse-grained allocation and periodic re-determination, with a task-based scheduler for fine-grained reallocation to address demand spikes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If constant dynamic independent allocation of each type of resource is performed, then resource utilization is optimized, but computational complexity becomes immense
Solution Approach 1:
The patent combines multiple independent resource allocation decisions into a single unified allocation decision. Instead of separately allocating CPU, memory, storage, and network resources, the system allocates a bundled resource package that contains all these resources together. This merging approach reduces computational complexity by eliminating the need for multiple independent allocation algorithms while maintaining efficient resource utilization through the bundled approach.
Solution Approach 2:
The unified resource allocation system serves multiple resource types simultaneously through a single allocation mechanism. The bundled resource package approach allows one allocation decision to satisfy multiple resource requirements (CPU, memory, storage, network) at once, making the allocation system universal and multi-functional rather than requiring separate specialized allocators for each resource type.
2Reliability
If resources are allocated to support peak demand, then reliability is improved, but resource idle time increases during non-peak periods
Solution Approach 1:
The system dynamically adjusts resource allocation by monitoring actual resource usage and reallocating bundled resources accordingly. During peak demand, resources are allocated to maintain service reliability, while during non-peak periods, the system identifies and reallocates idle bundled resources to other uses. This dynamic approach allows the system to maintain reliability when needed while minimizing resource idle time through periodic reevaluation and reallocation.
Solution Approach 2:
The system performs periodic reevaluation and reallocation of bundled resources at scheduled intervals. This periodic action allows resources to be allocated for peak demand periods while providing opportunities to identify and reallocate idle resources during non-peak periods, thus balancing reliability requirements with resource utilization efficiency over time.
3Measurement precision
If finest-grain scheduling is implemented, then resource allocation precision is improved, but system complexity increases
Solution Approach 1:
The patent merges the scheduling functions for multiple resource types into a single unified scheduler that operates at fine granularity. Instead of implementing separate complex schedulers for CPU, memory, storage, and network resources, the system uses one bundled resource scheduler that makes fine-grained allocation decisions for all resources simultaneously, reducing overall system complexity while maintaining allocation precision.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
An apparatus comprising a processor configured to obtain estimated processing requirements, estimated memory requirements, estimated storage requirements, and estimated network communication requirements for a plurality of data center (DC) tenants; determine a Minimum Resource Schedulable Unit (MRSU) for the tenants, the determined MRSU comprising a dynamically allocatable group of processor resources, processing memory resources, storage resources, and network resources comprised in at least one of a plurality of DC servers, wherein the MRSU is determined such that each DC tenant's estimated processing requirements, estimated memory requirements, estimated storage requirements, and estimated network communications requirements are met by allocation of a corresponding integer value of MRSUs; and allocate the corresponding integer value of MRSUs to each DC tenant as an MRSU allocation; and a transmitter coupled to the processor and configured to transmit the MRSU allocation to the DC servers for allocation to the DC tenants.