Distributed Virtual Switch Logic for VM Packet Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In NFV deployments, dedicating multiple physical cores to software virtual switches leads to bottlenecks, resulting in unacceptable switching latencies and throughput limitations, especially during peak packet traffic periods, and requires excessive packet classification, reducing flexibility and scalability.
Innovation Solution
The solution involves configuring a virtual switch by generating a global flow table and partitioning it into separate tables for each VM, allowing each VM to execute virtual switch logic and route packets independently, thereby reducing the load on the host operating system and processing elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple physical cores are dedicated to software virtual switches for routing packets between VMs, then packet routing capability is improved, but switching latency increases and throughput is limited during peak traffic periods
Solution Approach 1:
The patent segments the centralized virtual switch functionality into distributed virtual switch logic across multiple VMs. Each VM executes a portion of the virtual switch application, dividing the packet routing task into parallel segments that can be processed simultaneously, thereby reducing latency and increasing throughput without requiring dedicated physical cores for a single virtual switch instance.
Solution Approach 2:
The patent transitions from a single-dimension centralized virtual switch architecture to a multi-dimensional distributed architecture where virtual switch logic is spread across multiple VMs running on different processing elements. This dimensional shift enables parallel packet processing across multiple cores without the bottlenecks of centralized switching.
2Productivity
If multiple physical cores are dedicated to software virtual switches, then packet routing capability is improved, but device scalability is reduced due to bottlenecks
Solution Approach 1:
By segmenting the virtual switch application into distributed components across multiple VMs, the system can dynamically allocate and scale service chain resources. Each VM can independently handle portions of the service chain, allowing the system to scale by adding more VMs rather than being constrained by a fixed number of dedicated physical cores.
Solution Approach 2:
The distributed virtual switch architecture enables multiple service chain configurations to share the same infrastructure. VMs can be dynamically assigned different service chain roles based on demand, making the system universal and adaptable to various network service requirements without being locked into a fixed configuration.
3Measurement precision
If excessive packet classification is performed by the virtual switch, then packet routing accuracy is improved, but processing overhead increases and flexibility is reduced
Solution Approach 1:
The patent segments packet classification responsibilities across multiple distributed virtual switch instances running on different VMs. Each instance performs classification for its assigned packet flow, distributing the computational burden and reducing the complexity of any single virtual switch while maintaining overall classification accuracy through coordinated operation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Examples may include a host computing platform having a multi-core processor capable of supporting a host operating system and a plurality of virtual machines (VMs). Support may include processing elements or cores for the multi-core processor being capable of separately supporting individual VMs. The individual VMs may be capable of executing separate applications used to process packets for a service chain flow. In some examples, techniques for routing the packets for the service chain flow between the individual VMs may include distributing at least some switch logic between the individual VMs.