Commitment-Aware Scheduler for Cloud Workload Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed networks face challenges in efficiently scheduling workloads that are latency-sensitive versus latency-tolerant, as existing methods do not effectively balance cost and time-sensitivity, leading to suboptimal resource and cost utilization.
Innovation Solution
A scheduler in a distributed system determines whether to allocate workloads to dedicated or a combination of dedicated and non-dedicated computing resources, based on cost and time-sensitivity, using resource forecasting and priority levels to minimize committed resource usage while ensuring timely execution of latency-sensitive workloads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If dedicated computing resources are allocated in advance at a discounted rate, then cost is reduced, but resource availability and flexibility are limited
Solution Approach 1:
The system dynamically adjusts resource allocation by transitioning between committed and uncommitted resources based on real-time workload conditions. The scheduler monitors resource utilization and automatically shifts workloads between resource types, making the allocation dynamic rather than static, thus resolving the contradiction between cost efficiency and resource availability.
Solution Approach 2:
The system changes the parameter of resource allocation from fixed (committed) to variable (uncommitted) based on workload characteristics. By adjusting the allocation parameter dynamically according to demand, the system achieves both cost savings through committed resources and flexibility through uncommitted resources.
2Loss of time
If latency-sensitive workloads are prioritized for immediate execution, then time-sensitivity is improved, but resource consumption and cost increase
Solution Approach 1:
The system applies different quality levels of resource allocation to different workload types. Latency-sensitive workloads receive high-priority access to committed resources with guaranteed performance, while latency-tolerant workloads utilize uncommitted resources. This local differentiation resolves the contradiction by optimizing resource consumption for each workload category separately.
Solution Approach 2:
The resource pool is segmented into committed and uncommitted portions, and workloads are segmented into latency-sensitive and latency-tolerant categories. This segmentation allows the system to apply appropriate resource allocation strategies to each segment, ensuring time-sensitivity for critical workloads while controlling overall resource consumption.
3Loss of energy
If latency-tolerant workloads are delayed until dedicated resources are available, then cost is reduced, but productivity decreases
Solution Approach 1:
The system maintains continuous productive action by executing latency-tolerant workloads on uncommitted resources while committed resources are occupied, rather than idle-waiting. This continuity ensures productivity is maintained while still achieving cost reduction through strategic use of committed resources for time-critical tasks.
Solution Approach 2:
The system performs partial action by executing only the portion of workloads that can be handled by uncommitted resources immediately, while deferring the remainder to committed resources. This partial execution maintains productivity without incurring full cost penalties, resolving the contradiction between cost and productivity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system (100) including a distributed network of one or more virtual machines (166), having a first portion of committed virtual machines (168) dedicated to a user and a second portion of on-demand virtual machines (169). The system may also include a workload scheduler (184) configured to receive workloads (232) associated with the user. The scheduler may determine whether to schedule a given workload to be performed by a combination of virtual machines in the first and second portions, or by virtual machines included in only the first portion. The given workload may be scheduled to be performed by virtual machines in only the first portion at a first time if a sum of an anticipated resource consumption level of the given workload and a first consumption level of the first portion of virtual machines at the first time is less than or equals a total amount of resources included in the first portion.