Vector Bloom Filters for Network Flow Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional high-performance network routers using TCAMs for flow classification face challenges in power efficiency and size, while general-purpose computers struggle to match the performance of dedicated hardware in network functions like routing and packet filtering.
Innovation Solution
Implementing a system with multiple computing devices that use vector Bloom filters to classify network packets, allowing for high-probability matching of flow rules without masking headers, thereby reducing memory accesses and improving power efficiency and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If TCAM is used for flow classification, then classification performance is improved, but power consumption and device size increase
Solution Approach 1:
The patent segments the flow classification task into two phases: a fast path using vector Bloom filters for common packets, and a slow path using traditional methods for exceptional packets. This segmentation allows the system to achieve high overall performance while consuming less power, as most packets traverse the efficient fast path without requiring power-intensive TCAM operations.
Solution Approach 2:
The patent introduces vector Bloom filters as an intermediary data structure between the packet input and the flow table lookup. This intermediary enables probabilistic matching that significantly reduces the number of power-consuming memory accesses to the flow table, while still maintaining high classification accuracy through subsequent verification.
2Productivity
If TCAM is used for flow classification, then classification performance is improved, but device size increases
Solution Approach 1:
The patent uses vector Bloom filters as a compact copy or representation of the flow rules, storing only essential filtering information rather than complete flow entries. This copying approach achieves space-efficient storage that is dramatically smaller than TCAM while maintaining the ability to perform accurate flow classification through the two-phase lookup process.
Solution Approach 2:
The patent transitions from the traditional single-dimension approach of direct TCAM lookup to a multi-dimensional approach using vector Bloom filters that operate on multiple packet header fields simultaneously. This dimensional change enables more efficient use of storage space while maintaining classification capability through the vector-based filtering mechanism.
3Adaptability or versatility
If general-purpose computers are used for network functions, then cost and flexibility are improved, but performance compared to dedicated hardware deteriorates
Solution Approach 1:
The patent changes the operational parameters of general-purpose computers by implementing a two-phase classification system with vector Bloom filters. This parameter change transforms the performance characteristics, enabling software-based classification to achieve speeds and efficiencies previously only attainable with dedicated hardware, while retaining the flexibility of general-purpose computing platforms.
Data Source
AI summary
Technologies for efficient network flow classification include a computing device that receives a network packet that includes a header. The computing device generates a vector Bloom filter (VBF) key as a function of the header and searches multiple VBFs for a VBF that matches the VBF key. Each VBF is associated with a flow sub-table that includes one or more flow rules. Each flow sub-table is associated with a mask length. If a matching VBF is found, the computing device searches the corresponding flow sub-table for a flow rule that matches a masked header of the network packet. If no matching VBF is found or if no matching flow rule is found, the computing device searches all of the flow sub-tables for a flow rule that matches the header. The computing device applies a flow action of a matching flow rule. Other embodiments are described and claimed.