Asymmetric Network Address Encapsulation for Data Center Scalability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large-scale server virtualization in data centers leads to scalability issues with MAC forwarding table sizes due to flat MAC address spaces and increased broadcast traffic, particularly in Ethernet-based Layer 2 networks, where symmetric network address encapsulations fail to effectively reduce the number of entries in edge switches' forwarding tables without altering guest OS protocol stacks.
Innovation Solution
Implementing an asymmetric network address encapsulation scheme within a virtual switch of a physical server, where MAC-in-MAC encapsulation is performed, reducing the edge switch's forwarding table size by disabling MAC learning in the network and maintaining address decapsulation elsewhere, thus reducing the number of addresses maintained in intermediate switches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If symmetric network address encapsulation is used to reduce forwarding table size, then the number of FDB entries in core switches is reduced, but the FDB size in edge switches remains substantially large
Solution Approach 1:
The patent applies asymmetry by implementing unidirectional MAC address encapsulation where only ingress traffic undergoes encapsulation while egress traffic remains unencapsulated. This asymmetric approach reduces FDB entries in core switches without requiring edge switches to maintain large forwarding tables for all virtual machine MAC addresses, thereby resolving the contradiction between reducing overall forwarding table size and maintaining edge switch forwarding table manageability
2Reliability
If MAC learning is enabled in the network to maintain address mappings, then communication between virtual machines is maintained, but the forwarding table size in edge switches becomes substantially large
Solution Approach 1:
The patent segments the network into edge switches and core switches with different forwarding table functions. Edge switches maintain FDB entries only for directly connected physical servers, while core switches handle encapsulated MAC addresses for remote virtual machines. This segmentation allows reliable VM communication while keeping individual edge switch forwarding tables small and manageable
Solution Approach 2:
The patent introduces an intermediary encapsulation mechanism where ingress edge switches encapsulate source MAC addresses and egress edge switches perform decapsulation. This intermediary process enables edge switches to forward traffic based on physical server MAC addresses rather than maintaining extensive virtual machine MAC address tables, thus maintaining communication reliability while reducing forwarding table size
Data Source
Figure 23
Figure 24
Figure 25
AI summary
An apparatus comprising a service network, and a plurality of Layer 2 sites connected by the service network via a plurality of gateways, wherein the gateways are configured to map a plurality of Internet Protocol (IP) addresses of a plurality of hosts under a plurality of virtual local area networks (VLANs) in a plurality of Layer 2 sites to a plurality of addresses (e.g. MAC or others) of the corresponding gateways inform the other gateways in the other Layer 2 sites of the mapped IP addresses under each of the VLANs in the local Layer 2 sites, and forward data frames originated from the hosts in the local Layer 2 sites to the other gateways in the other Layer 2 sites when the data frames' destinations are residing in the other Layer 2 sites.