Priority-Based Scheduling for Memory Separation Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In datacenter servers, the current disaggregation technology struggles to maintain high bandwidth and low latency for memory connections, leading to performance degradation, especially when using power switches, and there is an imbalance in resource usage, with only a small percentage of applications utilizing a significant portion of resources, resulting in inefficient resource utilization.
Innovation Solution
A priority-based scheduling method that uses a host in a memory separation network to monitor traffic loads, update throttle values, and adjust bandwidths based on priority queues to prevent high-priority traffic from being blocked, ensuring that lower-priority traffic is not blocked first by differentiating throttle values and scheduling requests accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If memory disaggregation technology using optical switch is implemented to maintain high bandwidth and low latency, then connection performance between host and distributed memory is improved, but device complexity and cost increase
Solution Approach 1:
The patent segments memory resources into distributed memory pools that can be independently managed and allocated to different hosts. This segmentation allows the system to achieve high performance for critical applications while distributing the complexity across multiple manageable units rather than requiring a single complex optical switching fabric for all memory access.
2Productivity
If resources are distributed to improve resource utilization, then resource usage efficiency is improved, but connection bandwidth requirements increase
Solution Approach 1:
The patent implements priority-based scheduling that provides different quality levels of service to different traffic types. High-priority traffic receives guaranteed bandwidth and low latency service, while lower-priority traffic can utilize available capacity. This local quality differentiation allows efficient resource distribution without requiring all connections to have maximum bandwidth capacity simultaneously.
Solution Approach 2:
The scheduling mechanism dynamically adjusts resource allocation based on current system conditions and traffic priorities. When high-priority traffic is detected, the system dynamically reserves bandwidth and adjusts scheduling parameters to ensure performance requirements are met, while allowing more aggressive resource sharing during periods of lower demand.
3Reliability
If priority-based scheduling with differentiated throttle values is implemented, then high-priority traffic performance is maintained, but scheduling complexity increases
Solution Approach 1:
The patent changes the parameter of throttle values to be priority-dependent, where different priority levels have different default throttle values. This parameter differentiation allows the system to automatically provide appropriate service levels without complex real-time calculations, as the priority-based throttle values are pre-configured and applied based on traffic classification.
Data Source
AI summary
A priority-based scheduling method and a scheduler configured to perform the method are disclosed. The priority-based scheduling method performed by a host of a memory separation network includes receiving a traffic load of queues divided according to priorities from a distributed memory configured to perform a load monitoring function, updating a throttle value of the queues divided according to the priorities at regular intervals, selecting a queue to transmit a request to the distributed memory from among all the queues divided according to the priorities, and adjusting a bandwidth of the selected queue by using the throttle value when the request of the selected queue is transmitted to the distributed memory.