Bandwidth Allocation in Multi-Tenant Cloud Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cloud computing platforms lack network performance guarantees, leading to unpredictable application execution times and costs due to oversubscribed network resources, where bandwidth allocation is not effectively managed, affecting reliability and tenant costs in multi-tenant environments.
Innovation Solution
A method for determining optimal network bandwidth allocation in multi-tenant cloud networks by generating tree data structures representing virtual machines and network devices, with computed bandwidth values and weighted distance calculations to reserve the minimal required bandwidth, allowing for flexible and real-time adjustments based on tenant requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If network resources are oversubscribed to maximize resource utilization, then resource utilization improves, but network performance predictability deteriorates
Solution Approach 1:
The patent segments network resources by creating virtual network entities (VNETs) that provide dedicated network paths and bandwidth guarantees to different tenants. This segmentation allows oversubscription at the physical layer while maintaining predictable performance through logical isolation and reserved bandwidth allocation in virtual network layers.
Solution Approach 2:
The patent implements preliminary action by pre-allocating and reserving network bandwidth resources for tenants before actual network traffic occurs. Bandwidth reservations are established in advance through service level agreements, ensuring predictable network performance even when physical network resources are oversubscribed.
2Reliability
If bandwidth is reserved for tenants to ensure performance guarantees, then network performance predictability improves, but resource utilization deteriorates
Solution Approach 1:
The patent applies nesting by creating hierarchical virtual network structures where VNETs are nested within the physical network infrastructure. Multiple layers of virtualization (VNETs, virtual switches, virtual routers) are nested to provide bandwidth guarantees to tenants while allowing the underlying physical network to be oversubscribed and efficiently shared among multiple tenants.
Solution Approach 2:
The patent changes network parameters by introducing virtual network parameters (logical bandwidth, virtual paths) that are decoupled from physical network parameters. This allows bandwidth reservations to be configured in virtual space without directly constraining physical resource allocation, enabling both performance guarantees and efficient resource utilization through dynamic parameter adjustment.
3Ease of manufacture
If host-based virtualization is used to separate VMs, then implementation simplicity improves, but network performance isolation deteriorates
Solution Approach 1:
The patent introduces VNETs as intermediary entities between VMs and the physical network infrastructure. These virtual network entities act as mediators that provide enhanced network performance isolation by creating dedicated logical paths and enforcing bandwidth reservations, while building upon the existing host-based virtualization layer rather than replacing it.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computing device determines an optimal allocation of network bandwidth on a plurality of communications links for a plurality of virtual machines (VMs) in a multi-tenant cloud network based upon a set of bandwidth requirements for the plurality of VMs. The computing device generates a plurality of spanning tree data structures (TDSs). Edges in the TDSs include computed bandwidth values indicating an amount of bandwidth to be reserved over a communications link represented by that edge for the plurality of VMs. The computing device generates a weighted distance value for each TDS of the plurality of TDSs, and selects a TDS of the plurality of TDSs having a smallest weighted distance value as the TDS having computed bandwidth values representing the optimal allocation of network bandwidth.