Hardware-Assisted Transport Isolation for Fair NIC Resource Sharing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network communication systems face performance unfairness and slowdowns due to oversubscription of shared hardware resources, where slower connections consume resources intended for faster connections, leading to potential system crashes and inefficient resource utilization.
Innovation Solution
Implementing a hardware-assisted transport layer with dynamic resource limits that individually manage and pause connections exceeding resource thresholds, ensuring fair resource allocation and preventing global pauses by using a network interface card (NIC) to compute and update resource limits based on available resources and network congestion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple connections share hardware resources, then resource utilization increases, but slower connections cause head-of-line blocking that degrades performance of faster connections
Solution Approach 1:
The patent segments the shared hardware resources by creating separate resource pools: guaranteed resources allocated to individual connections and shared resources available to all connections. This segmentation prevents slower connections from blocking faster connections by ensuring each connection has dedicated resources while allowing fair sharing of additional resources.
Solution Approach 2:
The patent applies local quality by providing different resource allocation mechanisms to different connections based on their needs. Each connection receives guaranteed resources to maintain baseline performance, while faster connections can utilize additional shared resources. This creates localized quality differences where each connection experiences performance appropriate to its allocation rather than being uniformly affected by head-of-line blocking.
2Reliability
If the transport layer initiates a global pause to prevent system crashes, then resource exhaustion is avoided, but performance fairness between connections is lost and faster connections are impeded
Solution Approach 1:
The patent segments the pause mechanism from global to connection-level by implementing individual backpressure control for each connection. Instead of pausing all connections globally when resources are exhausted, the system applies backpressure only to the specific connection causing resource exhaustion. This maintains system stability while preserving performance fairness for other connections.
Solution Approach 2:
The patent introduces an intermediary resource allocation layer between the transport layer and hardware resources. This intermediary manages resource allocation dynamically, tracking guaranteed versus shared resource usage per connection and applying selective backpressure. This intermediary prevents global pauses by mediating resource access at the connection level, maintaining both system stability and performance fairness.
3Productivity
If dynamic resource limits are applied per connection, then performance fairness is improved, but device complexity increases due to additional resource management overhead
Solution Approach 1:
The patent segments resource management into distinct guaranteed and shared portions, simplifying the complexity by providing clear allocation rules. Each connection receives guaranteed resources upfront, and dynamic management is only required for shared resources. This segmentation reduces management complexity compared to fully dynamic allocation while maintaining performance fairness.
Solution Approach 2:
The patent implements self-service mechanisms where connections effectively manage their own resource usage within allocated limits. The backpressure mechanism allows connections to self-regulate by pausing transmission when they exceed their fair share of shared resources, reducing the need for complex centralized resource management while maintaining performance fairness.
Data Source
AI summary
Methods, systems, and apparatus, including computer-readable storage media for resource isolation between connections with shared hardware resources. A network device, such as a network interface card, is configured to determine dynamic resource limits for each connection, and backpressure each connection individually to avoid a global pause when the shared hardware resources are oversubscribed by the current connections. As a result, slower connections may be paused for exceeding resource limits, protecting faster connections from slowing down because resources are shared between both types of connections. Dynamic resource limits can be generated and updated not only per connection, but also based on subsets of the shared hardware resources assigned to different sources of data, as well assigned to different types of transactions communicated over a connection. A hardware-assisted transport layer can be configured to apply dynamic resource limits individually to different connections. from a variety of different upper-layer protocols (ULPs).