Load Balancer Bandwidth Isolation via Packet Dropping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional load balancers in cloud-computing networks randomly drop data packets across all VMs during high load, leading to unfair distribution of network capacity and performance degradation for unrelated tenants sharing physical resources.
Innovation Solution
Reprogramming load balancers to monitor bandwidth usage and probabilistically discard data packets based on the network's overall stress level and individual VM traffic, ensuring that endpoints with high loads are primarily affected while maintaining performance for those with low bandwidth usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional load balancers randomly drop data packets across all VMs during high load, then the network capacity is protected from overload, but unrelated tenants experience performance degradation and network capacity is unfairly distributed
Solution Approach 1:
The patent applies local quality by making different parts of the system (tenants/VMs) experience different packet dropping rates based on their individual bandwidth consumption. High-bandwidth consumers experience higher dropping rates while low-bandwidth consumers experience lower dropping rates, creating localized quality differences rather than uniform treatment across all tenants.
Solution Approach 2:
The patent changes the parameter of packet dropping probability from a uniform random value to a dynamic value that varies based on each tenant's bandwidth consumption. The load balancer monitors bandwidth usage and adjusts the dropping probability parameter accordingly, transforming a static random dropping mechanism into a dynamic, tenant-specific mechanism.
2Ease of operation
If load balancers monitor bandwidth usage and probabilistically discard packets based on individual VM traffic, then fair distribution of network capacity is achieved, but device complexity increases
Solution Approach 1:
The patent implements feedback by having the load balancer continuously monitor bandwidth consumption by each tenant and use this information to dynamically adjust packet dropping probabilities. The system feeds back the bandwidth usage data to the load balancing decision-making process, creating a closed-loop control system that adapts to changing network conditions.
Solution Approach 2:
The patent introduces an intermediary mechanism (the load balancer's bandwidth monitoring and decision-making system) that mediates between the network capacity constraints and the multiple tenants' bandwidth requirements. This intermediary processes bandwidth information and makes intelligent packet dropping decisions to achieve fair distribution without requiring direct negotiation or complex coordination between tenants.
Data Source
AI summary
Computerized methods, systems, and computer-storage media for fairly sharing available bandwidth among endpoints (including physical or virtual machines) of a multitenant environment are provided. Initially, a load balancer of the network is programmed to monitor bandwidth used by the endpoints, respectively, and to compare an aggregation of the usage against a network threshold. Also, the load balancer is equipped to dynamically classify the endpoints using the respective monitored bandwidth. Classifying generally involves designating some endpoints as small entities, which individually consume an amount of bandwidth less than an endpoint threshold, and designating other endpoints as large entities, which individually consume an amount of bandwidth greater than the endpoint threshold. Upon the aggregate bandwidth usage overcoming the network threshold, data packets being sent to or transmitted from the large entities are randomly dropped, while data packets being sent to or transmitted from the small entities are unrestrained.