Browse Prior Art Database

System and method to boost the efficiency of cache invalidation through machine learning in time manner Disclosure Number: IPCOM000248934D
Publication Date: 2017-Jan-22
Document File: 4 page(s) / 95K

Publishing Venue

The Prior Art Database


This document describes a method and system to reduce the volumn of cache invalidation messages in distributed systems. Tranditional cache invalidation message in single node system are more tightly related to the implementation of the each system which usually exposing a lot of schema details of the data structure. Invalidation to the cache data are triggered by business user operations and messages for invalidation are directly mapped to those data model details. From top level point of view, any bunch of data changes that tiggered by business users is related to each other. And from time manner, the data relationship could be identified by machine learning algroithm in history data. Classify algorithm in machine learning is used to help grouping the data and later help to compress the message volumn that sent through the whole messaging system.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 54% of the total text.


System and method to boost the efficiency of cache invalidation through machine learning in time manner

Business users doing large mount data updating in distributed services will generate a BIG volume of data synchronization events in side of local cache containers. Those events will flood the messaging queue that helping to transfer those update requests to remote services and make related cache data invalidated. This invention will help solve this large amount messages that flowing inside of the queues. The idea create a method/system by adding an extra message engine inside of the whole messaging systems to cluster related messages (similar and strong tied in time) and then instead of sending original invalidation messages, it sends out a group id tag which could be a replacement of original set of events. In this way, invalidation messages volume will drop down dramatically and we can remove the message chokes even business data changes intensively. 1. The description of original cache invalidation of original distributed services:

Previously: 1) Business user doing changes in source systems and source systems logging those changes and generate related invalidation messages and send

to message queues. 2) By monitoring related topics in message queues, target system will receive the invalidation message and drop/remove related stale data from local

data cache.

2. The proposed system:


1) Similar to above, the message will be sent to the message pattern process engine...