Hitless Packet Rule Upgrade via Merged TCAM Lists
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for updating packet processing rules in network devices, such as the make-before-break approach, face challenges in reducing storage requirements and ensuring hitless upgrades, especially when TCAM utilization exceeds 50%, leading to increased costs and resource constraints.
Innovation Solution
Creating a merged list of rules that includes the longest mergeable subsequence and rules unique to each list, which reduces storage needs by incorporating the overlap between old and new rules, allowing for a hitless upgrade by storing fewer rules in the TCAM.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the make-before-break approach is used to update packet processing rules, then hitless upgrades are achieved, but storage requirements in TCAM increase significantly
Solution Approach 1:
The patent merges the old rule set and new rule set into a single consolidated rule set that is installed in the TCAM. This merged rule set includes rules from both the old and new sets, along with transition rules that enable hitless upgrading. By merging rather than storing separate old and new rule sets, the patent reduces overall storage requirements while maintaining the ability to perform hitless upgrades.
Solution Approach 2:
The patent segments the rule set into different categories: rules that exist in both old and new sets, rules unique to the old set, rules unique to the new set, and transition rules. This segmentation allows for optimized storage by only storing the necessary rules in the TCAM and managing rule transitions through software-based logic, thereby reducing TCAM storage requirements while maintaining hitless upgrade capability.
2Productivity
If TCAM utilization exceeds 50%, then network device performance is maintained, but costs and resource constraints increase
Solution Approach 1:
The patent extracts certain rules from the TCAM storage and manages them through software-based rule sets. Specifically, rules that are common to both old and new rule sets, as well as transition rules, are managed in software rather than being stored entirely in the TCAM. This extraction reduces TCAM utilization and associated costs while maintaining packet processing throughput by leveraging the high-speed matching capability of the TCAM for critical rules.
3Ease of manufacture
If conventional update methods are used, then rule updates are simple, but ongoing packet processing is impacted
Solution Approach 1:
The patent performs preliminary actions by pre-installing the merged rule set in the TCAM before the actual rule update is needed. This merged rule set is prepared in advance and includes all necessary rules for both the old and new configurations, along with transition rules. When an update is triggered, the system can switch to the new rule set without impacting ongoing packet processing, as the infrastructure is already in place.
4Quantity of substance
If the merged list of rules is installed in overlapping memory portions, then storage space is optimized, but installation complexity increases
Solution Approach 1:
The patent introduces dynamic rule indicators (such as version bits or priority flags) that allow the system to dynamically manage which rules are active during the transition from old to new rule sets. This dynamic approach enables the merged rule set to be installed in overlapping memory portions without excessive complexity, as the system can flexibly activate or deactivate specific rules based on the upgrade state, simplifying the installation process while optimizing storage space.
Data Source
AI summary
Upgrading packet processing rules in a network device with a replacement set of rules includes generating an edit sequence that represents edit operations to transform an already-installed old set of rules into the replacement rules. The edit sequence is used to identify a subsequence of rules that is common to both the old rules and the replacement rules. A merged list is generated by a combination of the old rules, the replacement rules, and the common subsequence of rules. The merged list is downloaded to the network device, overwriting the old rules in bottom-up fashion allowing packet processing to continue concurrently using the old rules.