Resource-Centric Memory Address Allocation for Dynamic Sharing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing server-centric networks face challenges in efficiently managing asymmetric resource utilization, leading to underutilization of network resources due to the commercialization of 5G services and the shift towards memory-centric or computing-centric services, where CPUs or memories are underutilized despite available resources.
Innovation Solution
A resource-centric network architecture that allows multiple computing devices to share I/O devices as a large network virtual I/O device, managing memory pools by allocating and deleting memory areas dynamically, and enabling byte-access through individual offset values, with a memory address allocation apparatus that adjusts memory allocation based on optical link availability and capacity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If memory is pre-allocated in a server-centric network, then hardware implementation is simple and connection between devices is easy, but network resources cannot be efficiently shared and asymmetric utilization occurs
Solution Approach 1:
The patent segments the memory pool into multiple memory areas that can be dynamically allocated to different computing devices. Instead of dedicating entire memory resources to single servers, the memory is divided into allocable units that can be reassigned based on service requirements, enabling efficient resource sharing while maintaining manageable allocation granularity
Solution Approach 2:
The patent implements dynamic memory allocation where memory areas can be allocated and deallocated in real-time based on service demands. The network virtual I/O device continuously monitors resource utilization and adjusts memory allocation accordingly, transforming the static pre-allocated memory model into a dynamic shared resource model that adapts to changing service requirements
2Ease of operation
If memory is pre-allocated to each server, then device connection is straightforward, but resources remain unavailable when not fully utilized
Solution Approach 1:
The patent creates a universal network virtual I/O device that serves multiple computing devices simultaneously. This single virtual device manages the shared memory pool and can allocate memory to any connected computing device based on current service needs, making the memory resources universally accessible rather than dedicated to specific servers
Solution Approach 2:
The network virtual I/O device automatically monitors resource utilization and performs memory allocation without manual intervention. When a computing device requires memory, the system self-manages the allocation process by identifying available memory areas and assigning them to the requesting device, eliminating the need for complex manual resource management
3Adaptability or versatility
If memory areas are dynamically allocated, then resource sharing efficiency improves, but memory management complexity increases
Solution Approach 1:
The patent introduces a network virtual I/O device as an intermediary between computing devices and the physical memory pool. This intermediary layer abstracts the complexity of memory management by handling all allocation, deallocation, and monitoring operations centrally, allowing computing devices to access memory through simplified interfaces without directly managing the underlying complexity
Data Source
AI summary
A method and apparatus for allocating a memory address in a resource-centric network are disclosed. The method of allocating a memory address may include receiving a request for a new service, determining whether the new service is able to be accommodated in a virtual memory that is pre-allocated in a resource-centric network, when the new service is able to be accommodated, allocating a memory area for accommodating the new service to the virtual memory, and when the new service is not able to be accommodated, allocating the memory area by using an additionally allocated area of a virtual memory of the resource-centric network.