Adaptive Rate Limiters for VM Network Traffic Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In cloud computing environments, regulating network traffic to prevent disproportionate bandwidth consumption and ensure predictable and quantifiable networking performance is challenging due to issues like reachability isolation inadequacy, TCP congestion control limitations, and the complexity of conventional Quality of Service mechanisms.
Innovation Solution
A system that manages message processors within a hypervisor, detects congestion by calculating combined bandwidth, identifies violating message processors, and uses adaptive rate limiters to reduce bandwidth through a congestion-free network core, ensuring predictable and quantifiable networking performance by throttling VM-to-VM traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional QoS mechanisms (e.g., separate queues with Weighted Fair Queuing) are used to regulate network traffic, then network performance differentiation is achieved, but the system becomes complicated and expensive to use, especially when tenants frequently join and leave
Solution Approach 1:
The patent segments the network traffic regulation into two distinct components: a congestion-free network core that handles packet forwarding without queuing, and adaptive rate limiters at the edges that enforce bandwidth subscriptions. This segmentation eliminates the need for complex in-network QoS mechanisms while maintaining performance differentiation through simple per-flow rate limiting at the network boundaries.
Solution Approach 2:
The patent extracts the QoS enforcement function from the network core and places it at the network edges through adaptive rate limiters. This extraction removes the complexity of in-network queue management and scheduling from the core switching fabric, allowing the core to operate simply as a high-speed packet forwarder while QoS policies are applied at the periphery.
2Reliability
If TCP congestion control is used to isolate tenant performance, then some level of traffic regulation is achieved, but tenants can still achieve unbounded utilization by using many TCP flows and protocol variations
Solution Approach 1:
The patent implements a universal rate limiting mechanism that operates independently of the transport protocol being used. The adaptive rate limiters enforce bandwidth subscriptions at the network layer, applying uniformly to TCP, UDP, and any other protocols. This protocol-agnostic approach prevents tenants from evading rate limits by switching protocols or using multiple flow variations.
Solution Approach 2:
The patent introduces adaptive rate limiters as intermediary devices between the network core and tenant workloads. These intermediaries monitor and control the aggregate bandwidth consumption of multiple flows from a single tenant, acting as a mediator that enforces subscription limits regardless of how the tenant distributes traffic across multiple TCP connections or protocol variations.
3Reliability
If statically throttling each VM on the sender side is used to prevent bandwidth abuse, then bandwidth consumption is limited, but unused capacity is wasted and a tenant can still cause performance interference at any static rate applied
Solution Approach 1:
The patent replaces static rate limiting with dynamic adaptive rate limiting. The rate limiters continuously monitor the aggregate bandwidth consumption of all flows from a tenant and adjust the limiting threshold dynamically based on the tenant's subscription and current network conditions. This dynamic approach allows tenants to utilize unused capacity when available while preventing abuse when limits are approached, maximizing network efficiency.
Solution Approach 2:
The patent implements feedback mechanisms where adaptive rate limiters monitor actual bandwidth consumption and adjust rate limiting decisions in real-time. The system receives feedback about flow rates and aggregate consumption, then dynamically adjusts the applied rate limits to match subscription agreements and prevent interference. This feedback loop enables efficient capacity utilization while maintaining reliable bandwidth control.
Data Source
AI summary
The ensuring of predictable and quantifiable networking performance includes adaptively throttling the rate of VM-to-VM traffic flow. A receiving hypervisor can detect congestion and communicate messages for throttling traffic flow to reduce congestion at the receiving hypervisor.