In-Memory Database Result Persistence via Disk Caching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In-memory databases face challenges when the amount of memory required to calculate all algorithms across all versions exceeds available resources, leading to the need for evicting results to accommodate newer calculations.
Innovation Solution
A computer-implemented method that determines whether a result of a deterministic process exists in-memory, and if not, gathers dependencies, hashes them, and checks if the hash exists on disk. If it does, the results are retrieved from disk; otherwise, the process is triggered to generate results, with the results stored in-memory and/or on-disk.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If all algorithm results are stored in-memory, then access speed is improved, but memory usage increases and may exceed available resources
Solution Approach 1:
The patent segments the storage system into two distinct parts: in-memory storage for frequently accessed results and disk storage for less frequently accessed or larger result sets. This segmentation allows the system to maintain fast access for hot data while accommodating large volumes of data on disk, resolving the contradiction between access speed and memory usage constraints.
Solution Approach 2:
The patent transitions from a single-dimension storage approach (only in-memory) to a multi-dimensional storage hierarchy that includes both in-memory and disk storage layers. This dimensional expansion allows the system to optimize for both speed (in-memory layer) and capacity (disk layer), effectively resolving the resource constraint contradiction.
2Quantity of substance
If results are evicted from in-memory to make room for newer results, then memory usage is controlled, but access speed decreases due to recalculations
Solution Approach 1:
The patent implements preliminary action by pre-computing algorithm results and storing them on disk before they are needed. When results are requested, the system first checks in-memory cache, then retrieves from disk if necessary, avoiding the need to re-execute the algorithms. This preliminary storage of results eliminates the productivity loss that would occur from recalculating evicted results.
Solution Approach 2:
The patent introduces an intermediary disk storage layer between the in-memory cache and the algorithm execution engine. This intermediary serves as a persistent repository that can supply results to both the in-memory cache and direct query interfaces, reducing the frequency of algorithm re-execution and maintaining productivity while controlling memory usage.
3Reliability
If in-memory caches are invalidated based on data changes, then consistency is maintained, but additional processing overhead is introduced
Solution Approach 1:
The patent implements a feedback mechanism where the system monitors data changes and automatically invalidates corresponding in-memory cache entries. This feedback loop ensures that when source data changes, the cached results are invalidated and will be recomputed or retrieved from disk on next access, maintaining consistency while providing a systematic approach to cache management that balances overhead and reliability.
Data Source
AI summary
Systems and methods disclose herein procedures for accelerated tree learning. In one class, the acceleration is based on self-adapting learning rates, while in another class, the acceleration is based on a plurality of learning rates, wherein each learning rate varies over the training; each learning rate increases linearly as a respective pseudo residual maintains a direction across sequential training iterations; and each learning rate decreases exponentially as the respective pseudo residual changes direction across sequential training iterations. The latter can be incorporated with other methodologies, such as momentum-augmented gradient boosting and Nesterov Accelerated Gradient Boosting. These systems and methods for accelerated tree learning exhibit a marked reduction in training time and resources required for gradient boosted trees.