Memory Pool Configuration for Distributed Network Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The availability of memory resources in computing frameworks is constrained by physical limitations and costs, and existing solutions like increasing RAM do not effectively address memory constraints, especially due to issues like garbage collection, which can lead to increased job completion times and heat dissipation problems.
Innovation Solution
A memory pool or 'memory lake' configuration using non-volatile memory express (NVMe) and solid-state disks (SSDs) allows dynamic allocation and management of memory across computing nodes, providing a large-scale RAM-compatible resource that can be shared, reducing the need for hardware upgrades and optimizing memory access through a centralized controller.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If RAM footprint is increased to address memory constraints, then memory availability is improved, but cost and heat dissipation increase non-linearly
Solution Approach 1:
The patent creates a virtual copy of RAM functionality using SSD storage. The memory pool manager software creates a virtual memory file system that mimics RAM behavior by storing data blocks on SSD, allowing the system to access large amounts of memory without physically installing large amounts of RAM hardware, thus avoiding the associated heat and cost issues.
Solution Approach 2:
The patent changes the physical state and location of memory from volatile RAM to non-volatile SSD storage. By using NVMe SSDs with high I/O speeds, the system achieves memory-like performance characteristics while utilizing storage media that has better cost and power efficiency profiles compared to large-scale RAM configurations.
2Quantity of substance
If more RAM is provided to computing nodes, then memory capacity is improved, but garbage collection time increases causing system freezes
Solution Approach 1:
The patent introduces a memory pool manager software layer as an intermediary between the compute framework and the storage system. This manager handles memory allocation, data block management, and persistence operations, freeing the JVM from direct garbage collection responsibilities and allowing it to focus on computational tasks without frequent freezes.
Solution Approach 2:
The patent replaces the mechanical garbage collection process in RAM with a software-managed persistence system using SSD. Instead of relying on JVM garbage collectors to manage memory reclamation, the system uses the memory pool manager to handle data lifecycle management, leveraging the high I/O capabilities of NVMe SSDs to maintain performance.
3Quantity of substance
If hardware upgrades are performed to increase RAM, then memory resources are improved, but device complexity and cost increase
Solution Approach 1:
The patent makes the SSD storage system multi-functional by using it both for its traditional storage purpose and as a virtual RAM memory pool. The same NVMe SSD infrastructure serves dual purposes: persistent storage for data lakes and high-speed memory for compute operations, eliminating the need for separate RAM hardware upgrades.
Solution Approach 2:
The patent creates a virtual memory space that is a software copy of physical memory, implemented on top of SSD storage. This virtual memory file system provides the compute framework with access to large memory capacities without requiring physical RAM installation, simplifying hardware configuration while expanding memory resources.
Data Source
AI summary
An example operation may include one or more of allocating a memory node to a computing memory allocated to one or more computing nodes, allocating a software process dedicated to controlling the memory node, and modifying an existing file system of the one or more computing nodes to replace the existing file system with a new file system that includes the memory node.