Ethernet SAN Deterministic Flow Control via L2 FIP Protocol
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Ethernet Storage Area Networks (SANs) face challenges such as indeterminate link states, lack of deterministic flow control, and unreliable data transmission due to their reliance on best-effort delivery and congestion-induced frame losses, making them less suitable for sensitive data storage compared to Fibre Channel (FC) fabrics despite being more costly.
Innovation Solution
The implementation of a Layer-2 (L2) Fabric Initiation Protocol (FIP) domain and path exchange multicast frames, along with unicast frames, to establish a deterministic link topology and flow control within an Ethernet SAN, utilizing Software Defined Networking (SDN) and vendor-specific frames for link state awareness and error recovery, enabling reliable data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Ethernet fabrics are used instead of Fibre Channel fabrics, then cost is reduced, but reliability and determinism deteriorate due to best-effort delivery and lack of authentication mechanisms
Solution Approach 1:
The patent introduces an Ethernet fabric as an intermediary networking infrastructure that implements FC SAN functionality. By adding authentication mechanisms, link state detection, and flow control protocols to the Ethernet fabric, it mediates between the cost advantages of Ethernet and the reliability requirements of FC SANs, enabling reliable data transmission over Ethernet infrastructure.
2Device complexity
If conventional Ethernet link state detection mechanisms are used, then device complexity is reduced, but link state awareness speed deteriorates due to timeouts and keep-alive timers operating on seconds to minutes scale
Solution Approach 1:
The patent changes the time parameter of link state detection from seconds-to-minutes scale (conventional keep-alive timers) to a much faster scale by implementing continuous link state probes and hardware-level detection mechanisms. This parameter change enables rapid link state awareness while maintaining manageable device complexity through standardized protocol implementations.
3Ease of operation
If MAC PAUSE frames are used for flow control in Ethernet fabrics, then ease of operation is improved, but data loss increases during congestion due to frame discarding
Solution Approach 1:
The patent implements a feedback-based flow control mechanism using PFC (Priority Flow Control) that continuously monitors link congestion states and dynamically adjusts frame transmission rates. When congestion is detected, the system provides feedback signals to transmitting devices to slow down or pause transmission, preventing buffer overflow and frame discarding while maintaining ease of operation through automated control.
4Manufacturing precision
If DCB protocol standards are implemented for flow control, then manufacturing precision is improved with granular PFC mechanisms, but determinism and controllability still deteriorate compared to FC fabrics
Solution Approach 1:
The patent implements dynamic flow control mechanisms that adaptively adjust transmission parameters based on real-time link conditions. By combining PFC's granular priority-based control with dynamic congestion detection and adjustment algorithms, the system achieves deterministic flow control that responds adaptively to changing network conditions, surpassing static DCB implementations while maintaining manufacturing precision.
Data Source
AI summary
An Ethernet SAN system includes an initiator device and a target device coupled together by an Ethernet fabric that includes a plurality of Ethernet switch devices. The Ethernet switch devices exchange L2 FIP domain exchange multicast frames that include domain information associated with each of the Ethernet switch devices. That domain information is then used to exchange L2 FIP path exchange multicast frames that include path information associated with each of the Ethernet switch devices. That path information is then used to calculate path costs associated with at least some of the Ethernet switch devices. Those path costs are then used to exchange L2 FIP device exchange unicast frames that include device information associated with the initiator device and the target device. The Ethernet switch devices then transmit data between the initiator device and the target device based on the device information and the path costs.