End-Host Traffic Flow Control for Datacenter QoS Bottlenecks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Datacenters face challenges in providing fine-grained Quality of Service (QoS) and optimizing network resources due to the unique characteristics of their traffic flows, including bottlenecks and performance issues caused by long-lived and short-lived traffic flows sharing common bandwidth resources, which traditional traffic engineering approaches are ill-equipped to handle effectively.
Innovation Solution
A system comprising flow agents and a traffic engineering controller that identifies long-lived traffic flows at their source, allocates bandwidth based on QoS requirements, and controls the send rate to prevent bottlenecks, using a weighting function and augmenting function to maximize bandwidth utilization and ensure reliable QoS across the datacenter network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional traffic engineering approaches are used, then network resources are managed centrally, but fine-grained QoS cannot be provided and bottlenecks occur due to inability to control long-lived and short-lived traffic flows effectively
Solution Approach 1:
The patent segments traffic control by classifying traffic flows into long-lived and short-lived categories, and further segments them into priority classes (e.g., real-time, bulk transfer, background). This segmentation enables differentiated QoS treatment for different traffic types without requiring complex centralized control, resolving the contradiction between QoS reliability and control complexity.
Solution Approach 2:
The patent implements local quality control by enabling end hosts to autonomously identify their own traffic flows and apply local rate limiting based on globally communicated bandwidth allocations. This localizes QoS control to individual hosts rather than requiring complex centralized management, improving QoS reliability while reducing overall system complexity.
2Productivity
If bandwidth is allocated to all traffic flows, then bandwidth utilization is maximized, but bottlenecks occur when long-lived and short-lived flows share common resources
Solution Approach 1:
The patent implements dynamic bandwidth allocation where the traffic engineering controller continuously monitors traffic conditions and adjusts bandwidth allocations dynamically. Long-lived flows receive allocations based on their priority and resource requirements, while short-lived flows get allocated bandwidth as it becomes available. This dynamic adjustment maximizes bandwidth utilization while preventing bottlenecks through adaptive resource distribution.
Solution Approach 2:
The patent ensures continuous useful action by maintaining persistent bandwidth allocations for long-lived priority traffic flows while allowing short-lived flows to utilize remaining capacity. The system continuously monitors and adjusts allocations to ensure that high-priority traffic maintains its service level while still maximizing overall bandwidth utilization, preventing bottlenecks through continuous optimization.
3Reliability
If centralized traffic control is implemented, then global bandwidth allocation is possible, but scalability is limited due to the complexity of managing all traffic flows centrally
Solution Approach 1:
The patent introduces end hosts as intermediaries between the traffic engineering controller and individual traffic flows. The controller communicates bandwidth allocations to end hosts, which then enforce rate limiting locally. This intermediary approach maintains global QoS control through the controller while enabling scalability by distributing the actual traffic management workload to individual hosts, eliminating the need for complex centralized management of every flow detail.
4Object-affected harmful factors
If rate limiting is applied to all traffic, then bandwidth bottlenecks are prevented, but network performance deteriorates due to over-constraint
Solution Approach 1:
The patent applies local quality control by implementing rate limiting only where necessary and tailored to individual traffic flow characteristics. End hosts identify their own traffic flows and apply rate limiting based on their allocated bandwidth and priority level. This localized approach prevents bottlenecks for high-priority traffic while allowing lower-priority traffic to utilize available capacity without over-constraint, maintaining overall network performance.
Solution Approach 2:
The patent changes the parameters of rate limiting by adjusting bandwidth allocations and rate limits dynamically based on traffic conditions, priority levels, and available resources. The system modifies rate limiting parameters rather than applying fixed constraints, allowing optimization of bottleneck prevention while maintaining network performance through adaptive parameter adjustment that responds to real-time traffic patterns.
Data Source
AI summary
Systems and methods are disclosed for traffic engineering and traffic flow control in a datacenter. Flow agents may be provided at hosts that may be operable to control a rate at which a traffic flow is sent from one point to another within a datacenter to facilitate differing levels of Quality of Service (QoS). The differing levels of QoS may be implemented by differing sizes of bandwidth allocations assigned remotely by a traffic engineering controller. The traffic engineering controller may distribute the differing allocations of bandwidth to the flow agents corresponding to the traffic flows to keep the traffic flow rates within the bandwidth allocation. An augmenting function may also be applied at the traffic engineering controller to maximize bandwidth utilization by checking for unutilized bandwidth and reallocating such bandwidth in accordance with one or more QoS.