Relaxed Reliable Datagram Transport for High Performance Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High performance computing clusters face challenges in achieving high bandwidth and low latency due to the limitations of existing networking protocols, which often require significant processor involvement and are not scalable with increasing connections, and struggle with packet drops and reordering.
Innovation Solution
The implementation of a Relaxed Reliable Datagram transport service that separates packet reordering and reliability operations, handling reliability at the network adapter device and reordering in software, allowing for efficient packet delivery and reordering without the need for extensive processor involvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional networking protocols are used to ensure reliable packet delivery, then packet reliability is improved, but processor involvement increases and latency increases
Solution Approach 1:
The patent extracts reliability operations from the host processor and places them in the network adapter device. The network adapter independently handles packet transmission, retransmission, and acknowledgment without requiring processor intervention, thereby maintaining reliability while reducing latency and processor involvement.
Solution Approach 2:
The network adapter device performs self-service for reliability management by autonomously tracking transmitted packets, detecting packet drops, initiating retransmissions, and managing acknowledgments. This self-service capability eliminates the need for host processor involvement in reliability operations.
2Reliability
If traditional networking protocols are used to manage connections, then packet delivery control is improved, but scalability deteriorates with increasing connections
Solution Approach 1:
The patent extracts connection management functionality from the host processor and implements it in the network adapter device. Each network adapter independently manages its own connection state and packet tracking, allowing the system to scale to numerous connections without increasing host processor burden.
Solution Approach 2:
The patent segments the reliability management function into individual network adapter devices, where each adapter independently handles its own packets and connections. This segmentation allows parallel processing of multiple connections without centralized coordination overhead, improving scalability.
3Reliability
If packet reordering is handled at the network adapter device, then reliability is improved, but device complexity increases
Solution Approach 1:
The patent extracts reordering operations from the host processor and implements them in the network adapter device's reorder buffer. The network adapter independently manages packet reordering based on sequence numbers, ensuring reliable in-order delivery without requiring host processor involvement.
4Loss of time
If processor involvement is reduced for high performance computing, then latency is reduced, but reliability management becomes more difficult
Solution Approach 1:
The network adapter device provides self-service for reliability management by autonomously handling packet tracking, retransmission, acknowledgment, and reordering without host processor involvement. This self-service capability maintains reliability while achieving low latency performance.
Solution Approach 2:
The network adapter device acts as an intermediary between the host processor and the network, independently managing all reliability operations. This intermediary role shields the host processor from complex reliability management tasks while ensuring reliable packet delivery.
Data Source
AI summary
Provided are systems and methods for connection establishment over a network that does not require a user application to establish an explicit connection with a target application. In some implementations, provided is an apparatus configured to communicate with a network and a host device. The apparatus may receive a message and destination information associated with the message from the host device. The apparatus may further determine, using the destination information, a transport context from a plurality of transport contexts. The transport context may include a state of a connection with a destination on the network. The destination on the network may be associated with the destination information.