Session-Aware Edge Datastore for Consistent Replica Switching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing edge computing systems face challenges in maintaining session consistency when clients interact with multiple replicas across different datacenters due to mobility or dynamic reallocation of functionality, leading to inconsistent data access and increased latency.
Innovation Solution
A session-aware datastore system that utilizes session tokens to manage data storage nodes, ensuring consistency by transferring only relevant data subsets and employing a reconciliation algorithm that groups related data accesses into sessions, reducing data transfer and latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If eventual consistent models are used to replicate data across edge nodes, then data propagation is simplified and background updates are enabled, but session consistency is broken when clients switch between edges or when functionality is dynamically reallocated
Solution Approach 1:
The system performs preliminary actions by capturing queries in a cache before client requests arrive at different replicas. When a client switches edges, the destination node already has the queried data ready, ensuring session consistency without complex real-time coordination. This preliminary caching resolves the contradiction by preparing data in advance rather than relying on eventual consistency during client interactions.
Solution Approach 2:
The patent introduces an intermediary mechanism (query cache) between the data storage and client access layers. This cache acts as a mediator that holds queried data locally at each edge node, allowing clients to access consistent data regardless of which edge node they connect to. This intermediary resolves the session consistency issue while maintaining the simplicity of eventual consistency replication in the background.
2Reliability
If all data records are transferred when a client switches between edge nodes, then session consistency is maintained, but data transfer volume increases significantly
Solution Approach 1:
The system extracts only the necessary subset of data from the full dataset by capturing and caching specific queries that are actually needed for client sessions. When transferring data between edge nodes, only the extracted relevant queries are transferred rather than all data records. This extraction principle reduces data transfer volume while maintaining session consistency for the actual client needs.
Solution Approach 2:
Instead of transferring all data records (excessive action), the system transfers only the partial subset of queries that are actually needed for session consistency. This partial action approach transfers minimal necessary data (captured queries) rather than complete datasets, significantly reducing transfer volume while providing sufficient consistency for client operations.
3Reliability
If data transfer is performed synchronously when clients switch edges, then session consistency is ensured, but latency increases due to transfer waiting time
Solution Approach 1:
The system performs preliminary data capture and caching actions before clients need to access data at new edge nodes. By pre-capturing queries and maintaining them in caches at various edge nodes, the system eliminates the need for synchronous data transfer at the moment of client switching. This preliminary action ensures consistency is already in place, reducing latency to near-zero when clients switch edges.
Solution Approach 2:
The patent implements beforehand cushioning by pre-positioning queried data in caches at multiple edge nodes before clients actually need to access them. This creates a cushion of pre-available data that protects against latency when clients switch between edges. The system anticipates potential client movements and prepares data in advance, cushioning against the latency that would otherwise occur during data transfer.
4Quantity of substance
If query caching is implemented at edge nodes, then data transfer is reduced by transferring only captured queries, but additional storage and processing requirements are introduced
Solution Approach 1:
The system applies partial action by caching only the specific subset of queries that are actually executed, rather than caching all possible data or maintaining complete datasets. This selective caching of captured queries reduces storage requirements compared to full data replication while still achieving significant data transfer reduction. The partial caching approach balances storage overhead against transfer savings.
Solution Approach 2:
The patent changes the parameter of data storage from storing complete datasets to storing only captured query results. This parameter change transforms the storage requirement from full data replication to minimal query caching, reducing both storage and processing complexity at edge nodes. The system changes what is stored (from all data to only queried data), thereby reducing the complexity burden while maintaining transfer optimization.
Data Source
AI summary
A method includes creating a session token including a session identifier and a current node, capturing a set of queries in a cache from a client device at the current node associated with the session identification, receiving a request from the client for data access at a second node wherein the request includes the session token, transferring a subset of data records from the current node to the second node, wherein the subset of data records is selected based on the session identifier and the set of queries; and granting access to the client to the second node.