Forwarding-Plane Consistent Hashing for Symmetric Traffic Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for managing hash tables in network devices fail to provide traffic flow consistency and symmetricity, leading to service and session disruptions, and consume excessive computing and networking resources.
Innovation Solution
Implementing symmetric consistent hashing in the forwarding plane by sorting hash keys based on next hop addresses to generate a sorted hash table, ensuring the same hash table index is used in both upstream and downstream directions, independent of routing protocols.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional hash table management techniques are used in network devices, then routing functionality is provided, but traffic flow consistency and symmetricity are not maintained, leading to service and session disruptions
Solution Approach 1:
The patent segments the hash table management into two distinct parts: (1) control plane operations that generate hash keys and routing decisions, and (2) data plane operations that use pre-computed sorted hash tables for forwarding. This segmentation allows the data plane to use simple, consistent hash lookups while the control plane handles the complexity of hash key generation and table construction, thereby maintaining traffic flow consistency without overwhelming device complexity.
Solution Approach 2:
The patent applies preliminary action by pre-computing and storing sorted hash tables in the data plane before actual traffic forwarding begins. The control plane generates hash keys, sorts them by next-hop address, and constructs the sorted hash tables in advance. When traffic arrives at the data plane, the forwarding decision is made immediately using the pre-sorted table, ensuring consistent and symmetric routing without real-time computation overhead.
2Productivity
If traditional hash table management techniques are used in network devices, then routing functionality is provided, but excessive computing and networking resources are consumed
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing sorted hash tables in the data plane before actual traffic forwarding begins. The control plane generates hash keys, sorts them by next-hop address, and constructs the sorted hash tables in advance. When traffic arrives at the data plane, the forwarding decision is made immediately using the pre-sorted table, ensuring consistent and symmetric routing without real-time computation overhead.
Solution Approach 2:
The patent uses copying by creating and storing duplicate sorted hash tables in the data plane that mirror the control plane's routing decisions. The data plane maintains local copies of the sorted hash tables, allowing independent and rapid forwarding decisions without repeatedly querying the control plane, thus reducing resource consumption while maintaining service reliability.
3Reliability
If hash keys are not sorted by next hop address, then simpler hash table generation is possible, but symmetric routing with other network devices cannot be achieved
Solution Approach 1:
The patent segments the hash table management into two distinct parts: (1) control plane operations that generate hash keys and routing decisions, and (2) data plane operations that use pre-computed sorted hash tables for forwarding. This segmentation allows the data plane to use simple, consistent hash lookups while the control plane handles the complexity of hash key generation and table construction, thereby maintaining traffic flow consistency without overwhelming device complexity.
Solution Approach 2:
The patent applies parameter changes by transforming the hash key structure to include sorted next-hop address fields. Instead of using arbitrary or simple hash keys, the control plane generates hash keys that are explicitly sorted by next-hop address, and the data plane constructs hash tables based on these sorted keys. This parameter transformation ensures that the same hash key produces the same table index in both upstream and downstream directions, achieving symmetric routing.
Data Source
AI summary
A network device may receive a hash key from a control plane of the network device, and may sort the hash key based on a next hop address field and to generate a sorted hash key. The network device may store the sorted hash key locally in a data plane of the network device, and may generate a sorted hash table based on the sorted hash key. The network device may utilize the sorted hash table to symmetrically route traffic with another network device.


