RDMA Remote Node Control for Serial-Free Data Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network storage systems require serial operations and high latency for data retrieval, involving significant CPU processing and network bandwidth due to the need for serial communication to locate and retrieve data across multiple nodes.
Innovation Solution
Implementing a delegation node that offloads CPU processing by using RDMA connections to directly control RDMA queues of serving nodes, allowing direct data transfer without serial communication, thus reducing latency and bandwidth usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional protocols with multiple IO queues are used, then parallelism and control over IO operations are improved, but CPU processing requirements and device complexity increase
Solution Approach 1:
The patent extracts the CPU processing burden by implementing RDMA (Remote Direct Memory Access) technology that enables direct memory access between nodes without CPU intervention. The RNIC (RDMA Network Interface Card) handles IO operations autonomously, removing the need for complex CPU-based protocol processing while maintaining parallelism through multiple RDMA queues.
Solution Approach 2:
The patent replaces the mechanical CPU-based IO processing system with an automated hardware-based RDMA system. The RNIC autonomously manages IO queues, data transfer, and protocol handling, substituting software-based CPU processing with hardware-based direct memory access operations that achieve parallelism without proportional increases in device complexity.
2Device complexity
If serial operations are used to locate and retrieve data across multiple nodes, then device complexity is reduced, but latency and processing time increase
Solution Approach 1:
The patent implements preliminary action by pre-establishing RDMA connections and configuring RNIC queues before data retrieval operations. The delegation node pre-loads data location information and prepares RDMA write queues in advance, enabling immediate data transfer when retrieval is requested, thereby eliminating serial lookup delays while maintaining simple device architecture.
Solution Approach 2:
The patent introduces a delegation node as an intermediary that coordinates between the initiator node and serving nodes. The delegation node's RNIC receives RDMA write requests, determines the appropriate serving node, and directly initiates data transfer using pre-configured RDMA queues, eliminating the need for serial communication protocols while keeping individual node complexity low.
3Measurement precision
If CPU involvement is used for data location and transfer, then measurement precision and control are improved, but processing time and energy consumption increase
Solution Approach 1:
The patent implements self-service by enabling the RNIC to autonomously handle data location and transfer operations without CPU involvement. The RNIC uses pre-configured queues, keys, and address information to independently manage RDMA write requests, data placement, and transfer coordination, achieving precise control while eliminating CPU processing delays and reducing energy consumption.
Solution Approach 2:
The patent extracts CPU involvement from the data retrieval process by delegating all IO operations to the RNIC. The RNIC autonomously interprets RDMA commands, manages write queues, and coordinates data transfer between nodes, removing the CPU from the critical path and enabling parallel processing without sacrificing control precision.
4Productivity
If network bandwidth is increased for data retrieval operations, then productivity is improved, but energy consumption and cost increase
Solution Approach 1:
The patent implements continuity of useful action by establishing persistent RDMA connections and pre-configuring write queues that remain active for multiple data transfer operations. Once an RDMA connection is established, data can be transferred continuously using the same hardware resources without repeated protocol handshakes or CPU interventions, achieving high throughput with minimal energy consumption per byte transferred.
Data Source
AI summary
A first node, a second node, and one or more third nodes for controlling a remote node via RDMA connection are provided. The first node is connected by RDMA connection to the second node and the third nodes, and the second node is connected by RDMA connection to the third nodes. The first node receives a read request for data from the second node, determines at which third node the data is available, and causes the third node to send the data to the second node. The third node is caused to send the data by receiving a RDMA write request or a send for a command from the first node. The third node sends the data by performing a RDMA write operation or a send operation in response to the command written by the first node directly to a send queue at the third node.