Gateway Router VM Mobility Routing Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In Multiprotocol Label Switching (MPLS) Ethernet virtual private networks, data center inbound routing becomes inefficient and scalable issues arise when virtual machines move across data centers, leading to sub-optimal traffic routing and scalability problems due to the need to inject host routes into the wide area network, which third-party networks may not support.
Innovation Solution
Implementing a computer-implemented method using a first gateway router that detects VM mobility and registers the new location with a Locator/ID Separation Protocol (LISP) mapping system, allowing traffic to be tunneled directly to the new data center without polluting wide area network routing protocols, and notifying the departing data center to clean up state information, thus avoiding sub-optimal detours.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If host routes of moved VMs are injected into the WAN BGP to prevent traffic tromboning, then routing optimization is achieved, but scalability issues arise and third party networks may refuse to accept such fine-grained host routes
Solution Approach 1:
The patent extracts the routing information for moved VMs from the WAN BGP system and places it in a local routing table at the gateway router. This allows the gateway to have immediate access to optimized routes for VMs that have moved to other data centers without injecting host-level routes into the WAN, thereby preventing traffic tromboning while avoiding scalability issues and third-party network restrictions.
Solution Approach 2:
The patent introduces an intermediary mechanism (the gateway router's local routing table) that mediates between the need for routing optimization and the constraints of WAN BGP scalability. The gateway router maintains local state about moved VMs and uses this information to optimize routing decisions without requiring fine-grained host routes to propagate through the entire WAN.
2Reliability
If host-level information is injected into the WAN to track VM mobility, then traffic routing is optimized, but an explosion of states occurs in data centers as the set of host route information maintained increases
Solution Approach 1:
The patent extracts only the necessary routing information for moved VMs from the WAN BGP system and maintains it locally at the gateway router. This prevents the explosion of host route information in data centers by keeping detailed VM location state localized rather than propagating it throughout the WAN, thus optimizing traffic routing without creating state explosion.
3Measurement precision
If fine-grained host routes are used to track VM location, then precise routing is achieved, but third party networks between data centers refuse to accept such routes
Solution Approach 1:
The patent extracts the precise VM location information from the WAN routing system and keeps it local at the gateway router. This allows the gateway to make precise routing decisions for moved VMs without requiring third-party networks to accept or propagate fine-grained host routes, thus achieving accurate VM location tracking while maintaining compatibility with third-party network constraints.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and apparatus for optimizing data center routing in the event of virtual machine (VM) mobility are provided. In one embodiment, a first gateway router, acting as an interface between an Ethernet Virtual Private Network (EVPN) domain and a Locator/ID Separation Protocol (LISP) domain, detects EVPN mobility messages advertised when a VM that has moved connects to a gateway router at a data center. The first gateway router then initiates a LISP mobility event that registers the new location of the moved VM to a LISP mapping system. In another embodiment, the first gateway router may notify a second gateway router, located at another data center from which the VM departed, to clean up the state maintained in that data center. This notification may be made via EVPN or LISP mechanisms. In response, the second gateway router may insert a new sequence into the other data center.