Prefix-based fat flows for cloud data center load balancing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional systems in cloud data centers face challenges in handling large numbers of flows due to resource limitations, particularly during events like DDOS attacks, where microflows do not scale effectively and disrupt load balancing applications.

Innovation Solution

The implementation of prefix-based fat flows, where virtual network controllers configure virtual routers to aggregate multiple microflows into a single fat flow based on source or destination IP addresses and port ignoring, using fat flow rules stored in M-trie data structures for efficient forwarding.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If microflows are used to handle network traffic, then flow management precision is improved, but system scalability deteriorates under high load conditions such as DDOS attacks

Engineering Contradiction:
Improveflow management precisionVSAvoidsystem scalability
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent merges multiple microflows into a single fat flow by aggregating flows that share common source or destination IP address prefixes. This is achieved by defining fat flow rules that group flows based on IP prefix matching, allowing the system to treat multiple individual flows as a single aggregate flow for routing and processing purposes, thereby reducing the total number of flow entries needed in the system

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent changes the flow differentiation parameters from using individual IP addresses and ports to using IP address prefixes. By transitioning from fine-grained microflow identification to coarser prefix-based aggregation, the system maintains sufficient routing precision while dramatically reducing the number of distinct flows that need to be managed, thereby improving scalability

Inventive Principle:
Principle #35Parameter changes

2Productivity

If the number of distinct flows is reduced through aggregation, then processing resource consumption is reduced, but flow differentiation precision may be compromised

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidflow differentiation precision
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by differentiating flows at different levels of granularity. At the prefix level, flows are aggregated for efficient processing, but within each prefix group, individual flows maintain their identity for purposes such as load balancing. This allows the system to achieve both efficiency through aggregation and precision where needed through selective differentiation based on local requirements

Inventive Principle:
Principle #3Local quality

3Productivity

If fat flow rules are implemented to aggregate flows, then load balancing performance is improved, but device complexity increases

Engineering Contradiction:
Improveload balancing performanceVSAvoidflow rule configuration complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements universality by designing fat flow rules that can aggregate flows based on multiple criteria including source IP prefixes, destination IP prefixes, and combinations thereof. A single fat flow rule structure serves multiple aggregation purposes and can be applied to different flow types, reducing the need for multiple specialized rule sets and simplifying the overall configuration approach

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP3734916A1Prefix-based fat flows
Publication Date: 2020.11.04 JUNIPER NETWORKS INC
  • EP3734916A1 patent drawingFigure 1
  • EP3734916A1 patent drawingFigure 2
  • EP3734916A1 patent drawingFigure 3

AI summary

A network device includes one or more processors configured to use a fat flow rule that specifies at least one of a mask to be applied to source Internet protocol (IP) addresses or to destination IP addresses, or that source ports or destination ports are to be ignored. The one or more processors may further be configured to receive packets having different source or destination IP addresses and/or different source or destination ports, and nevertheless assign the packets to the same fat flow according to the fat flow rule, e.g., by masking the source or destination IP addresses and/or ignoring the source or destination ports of the packets. In this manner, the network device may aggregate two or more different flows into a single fat flow.