SDN Compiler for Unified Network Storage Compute Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Software Defined Networking (SDN) approaches face challenges in achieving tight integration between networking, storage, and computing resources, leading to sub-optimal performance due to loose coupling and limited visibility between network and application layers, resulting in complex operations and sub-optimal software development.
Innovation Solution
A method and compiler that translate high-level network specifications into instructions for physical and virtual networking and computing resources, using matrices to represent logical and physical resources and their relationships, allowing for direct forwarding decisions without requiring conversion on physical nodes and eliminating the need for an underlay network, thereby enabling more complex and flexible network service implementation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional distributed networking control protocols are used, then network autonomy and device independence are maintained, but integration between networking, storage, and computing resources is loose and visibility is limited
Solution Approach 1:
The patent merges networking, storage, and computing control into a single centralized SDN controller that can program all three resource types through unified interfaces. This consolidation enables tight integration where the controller has global visibility and can coordinate operations across all resource types, eliminating the loose coupling of traditional distributed protocols while managing complexity centrally rather than at each device.
Solution Approach 2:
The SDN controller implements universal programming interfaces that can instruct packet forwarding devices, storage equipment, and computing equipment through standardized protocols. This multi-functionality allows a single control plane to manage diverse resource types with consistent methods, improving adaptability across networking, storage, and computing while simplifying the operational complexity through a unified approach.
2Loss of information
If centralized SDN controller is used to program all resources, then tight integration and global visibility are achieved, but control plane complexity increases
Solution Approach 1:
The patent introduces information models as intermediary layers between the control plane and the diverse networking, storage, and computing resources. These models provide standardized representations of resource states and capabilities, allowing the controller to maintain global visibility without directly managing the full complexity of each resource type. The models act as mediators that translate between heterogeneous resource interfaces and the unified control plane.
3Extent of automation
If forwarding tables are distributed to switches via SDN controller, then centralized control is achieved, but programming access to physical hosts, virtual hosts, and NICs is limited
Solution Approach 1:
The patent extends the SDN controller's programming capability to provide universal interfaces for instructing packet forwarding devices, storage equipment, and computing equipment including physical hosts, virtual hosts, and NICs. This multi-functional programming interface allows centralized automation while adapting to the specific requirements of each resource type through standardized protocols and information models.
4Ease of manufacture
If logical network specification is translated to physical node instructions, then network services can be implemented, but conversion requirements on physical nodes increase complexity
Solution Approach 1:
The patent extracts the complex translation and conversion logic from physical nodes and concentrates it in the SDN controller. The controller performs all necessary translations between logical network specifications and physical device instructions, then distributes simplified commands to physical nodes. This extraction eliminates the need for physical nodes to perform complex conversions, reducing their complexity while easing network service implementation through centralized processing.
Data Source
AI summary
Method of and a compiler for controlling a network based on a logical network model. The network has physical nodes and logical nodes. The physical nodes are interconnected by physical links in accordance with a physical network layout. The logical network model has logical nodes indicated with a logical node name which refers to at least one physical node in the network. The method uses a depth-mapping relation defining how the logical nodes are mapped to the physical nodes. The method includes creating logical links between the logical nodes in dependence on the physical paths between the physical nodes and on the depth-mapping relation. The method uses edge-relationships between logical link, logical path, physical link, physical path and depth-mapping relations. Logical paths in the logical network are transformed into a physical path comprising of physical links between the physical nodes through recursive calculation and forwarding instructions are created for the physical nodes, in dependence on the edge-relationships and point-of-attachment names between physical links and physical nodes. A user of a compiler may specify additional operations other than switching, multiplexing or de-multiplexing to be performed at a logical node on packet or signal. Said packet or signal may be identified with a logical identifier identifying at least one logical link or logical path, and said additional operation may be specified at a logical node, providing programmability of additional operations in said logical network model. Said additional operations will, if possible, be performed by physical or virtual resources represented by physical nodes.