Network Interface Device Offloading Proxy Operations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Service mesh sidecars in cloud environments consume system resources and introduce latency in service-to-service communications, as they require processor resources that could be used for application execution and involve complex network processing.
Innovation Solution
Offloading proxy operations to a network interface device with separate compute, networking, and memory resources, allowing host processors to focus on service workloads while utilizing the network interface device for packet processing, thereby reducing latency and improving communication quality of service.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If service mesh sidecars are deployed on host processors to mediate service-to-service communications, then service mesh functionality is provided, but CPU and memory resources are consumed and communication latency increases
Solution Approach 1:
The patent segments the service mesh proxy functionality from the host processor by deploying it on a separate network interface device. This physical segmentation allows the host processor to focus on application workloads while the network interface device handles service mesh communications, thereby reducing resource consumption and communication latency on the host system.
Solution Approach 2:
The patent extracts the proxy operations from the host processor and relocates them to a network interface device. This extraction removes the burden of service mesh communications from the host CPU and memory, freeing these resources for application execution while maintaining service mesh functionality through the dedicated network interface device.
2Reliability
If service mesh sidecars are deployed on host processors, then service-to-service communication control is achieved, but system resources that could be used for application execution are consumed
Solution Approach 1:
The patent segments system resources into two distinct domains: host processor resources dedicated to application execution, and network interface device resources dedicated to service mesh communications. This segmentation ensures that application resources are not consumed by service mesh overhead while maintaining full communication control capability.
Solution Approach 2:
The patent extracts proxy operations and their associated resource consumption from the host system and relocates them to a separate network interface device. This extraction preserves host system resources for application workloads while the network interface device independently handles service mesh communication control.
3Reliability
If proxy operations are executed on host processors, then service mesh functionality is provided, but device complexity increases
Solution Approach 1:
The patent segments the system into two functional domains: a simplified host system dedicated to application execution, and a dedicated network interface device that encapsulates all service mesh proxy functionality. This segmentation reduces host system complexity while maintaining complete service mesh capability through the specialized network interface device.
Solution Approach 2:
The patent extracts service mesh complexity from the host system and consolidates it into a separate network interface device. This extraction simplifies the host system architecture by removing proxy-related components and their configuration complexity, while the network interface device independently provides the required service mesh functionality.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Examples described herein relate to a system for offloading a proxy for microservice-to-microservice communication to a network interface device. In some examples, the system includes a host interface and a network interface device circuitry comprising circuitry coupled to the host interface. In some examples, the circuitry is configured to: perform offloaded proxy operations of a service mesh interface for multiple services, wherein the circuitry is accessible via a virtual network device by a host processor-executed service of the multiple services and wherein the service mesh interface is to provide access to a service mesh to communicate with one or more services.