RDMA NIC Multicast Data Distribution in Cloud Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multicast transmission in cloud-based networks is unreliable due to the inability to prevent certain host servers from receiving data and lacks scalability, while unicast protocols create bottlenecks at the sending host server by requiring individual data transmission and acknowledgement signals for each host server.
Innovation Solution
The implementation of a distributed data transfer method using remote direct memory access (RDMA) network interface cards (NICs) that generate work queue elements (WQEs) to efficiently transfer data between host devices, bypassing the core processing of intermediary devices and allowing for reliable, scalable data distribution through a chain or tree structure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multicast protocol is used for data transmission, then data can be sent to multiple host servers simultaneously, but the transmission becomes unreliable and non-scalable as the number of receivers increases
Solution Approach 1:
The patent segments the data distribution task by introducing intermediary host servers that form a hierarchical structure. The originating host server divides the multicast group into smaller subsets, with each intermediary server responsible for a specific subset. This segmentation allows the system to maintain reliability through selective acknowledgment while preserving efficiency through parallel distribution channels.
2Reliability
If unicast protocol is used for data transmission to selected host servers, then the sending host server can track which servers received data through acknowledgements, but it creates a bottleneck at the sending host server
Solution Approach 1:
The patent introduces intermediary host servers as mediators between the originating host server and the final receiving servers. These intermediaries receive data via unicast from the origin, then distribute it further using multicast mechanisms. The intermediaries handle acknowledgment tracking locally, relieving the originating server from the bottleneck of processing individual acknowledgments from all receivers while maintaining transmission reliability.
3Adaptability or versatility
If the number of host servers in multicast groups increases, then broader data distribution is achieved, but the ability to prevent certain servers from receiving data is lost
Solution Approach 1:
The patent applies local quality by assigning different distribution policies to different segments of the network. Each intermediary host server manages its own subset of receivers with specific access controls and distribution rules. This allows the originating server to maintain broad distribution coverage while each intermediary enforces local selective control over its assigned subset, preserving the ability to prevent specific servers from receiving data.
Data Source
AI summary
An originating host device in a distribution chain is provided upstream from multiple host devices including intermediary and terminating host devices. The originating host device includes a core with a generation application and a first RDMA NIC. The core: determines a plan for transferring data between the originating host device and the other host devices; and generates WQEs to implement the plan. The first RDMA NIC includes a read application, a descriptor application, and a reception application. The read application is configured such that the first RDMA NIC reads the WQEs from a first memory. The descriptor application is configured such that the first RDMA NIC: writes a portion of the WQEs directly from the first RDMA NIC to a second memory of the intermediary host device; and triggers a second RDMA NIC of the intermediary host device to process the portion of the WQEs stored in the second memory.