Dynamic Memory Allocation for Edge Computing Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Edge computing systems face challenges in managing data latency and memory bandwidth effectively, particularly in scenarios where multiple tenants with different service level agreements (SLAs) share resources, leading to increased total cost of ownership (TCO) and potential failure to meet latency and bandwidth requirements.
Innovation Solution
A device is configured to dynamically select a memory device or tier based on latency, memory bandwidth, and TCO requirements, using programmable circuitry to prioritize memory allocation and storage, allowing for conditional resource allocation and encryption parameters to optimize data processing and storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If data is transferred to remote computers in cloud computing, then computing and storage resources are improved, but latency increases making it unacceptable for time-sensitive applications
Solution Approach 1:
The patent segments computing resources into cloud computing (remote data centers) and edge computing (local base stations). Time-sensitive data is processed at the edge closer to the client, while non-time-sensitive data can be processed in the cloud. This segmentation resolves the contradiction by providing different processing locations for different latency requirements.
Solution Approach 2:
The patent introduces an intermediary mechanism (the edge computing system at base stations) between the client device and the remote cloud data centers. This intermediary processes data locally when low latency is required, eliminating the need to transfer all data to remote computers while still providing access to cloud resources when appropriate.
2Productivity
If multiple tenants share memory resources in edge computing systems, then resource utilization is improved, but it becomes difficult to meet diverse service level agreement requirements for different tenants
Solution Approach 1:
The patent implements dynamic memory allocation where the system can adjust memory bandwidth and access priorities in real-time based on current SLA requirements. Different tenants can receive different levels of memory resource allocation dynamically, allowing the system to meet diverse SLA requirements while maintaining high overall resource utilization through shared infrastructure.
Solution Approach 2:
The patent applies local quality by providing differentiated memory service qualities to different tenants based on their specific SLA requirements. Each tenant receives memory resources with properties tailored to their needs (e.g., guaranteed bandwidth for some, best-effort for others), allowing diverse SLA compliance while sharing the same physical memory infrastructure.
3Loss of time
If memory resources are allocated to meet latency requirements, then latency performance is improved, but total cost of ownership increases
Solution Approach 1:
The patent changes the parameter of memory allocation from static to dynamic, adjusting memory bandwidth and allocation based on actual latency requirements and priority levels. This allows the system to provide low-latency service only when necessary, reducing overall energy consumption and TCO while maintaining required performance levels for time-sensitive applications.
Data Source
AI summary
Examples described herein relate to a device configured to allocate memory resources for packets received by the network interface based on received configuration settings. In some examples, the device is a network interface. Received configuration settings can include one or more of: latency, memory bandwidth, timing of when the content is expected to be accessed, or encryption parameters. In some examples, memory resources include one or more of: a cache, a volatile memory device, a storage device, or persistent memory. In some examples, based on a configuration settings not being available, the network interface is to perform one or more of: dropping a received packet, store the received packet in a buffer that does not meet the configuration settings, or indicate an error. In some examples, configuration settings are conditional where the settings are applied if one or more conditions is met.