Any-Local-Quorum Consistency For Distributed Database Read Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing distributed database systems face challenges in ensuring data availability and efficiency in read-after-new-write operations, particularly due to inconsistencies between write and read consistency levels across multiple data centers.
Innovation Solution
The introduction of an any-local-quorum consistency level allows for faster and more efficient data operations by enabling a system to proceed with just one data center responding to a request based on its local quorum, while ensuring data availability by pairing it with appropriate consistency levels on opposing operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a quorum consistency level is used for read operations across all sites, then data availability is improved, but read operation time increases significantly
Solution Approach 1:
The patent segments the distributed database into multiple sites, each with its own local quorum. Instead of requiring a quorum across all sites for every read operation, the system allows read operations to be satisfied by a quorum at any single site. This segmentation enables faster reads (by querying only one site) while maintaining data availability (by ensuring the data exists in at least one site's quorum).
Solution Approach 2:
The patent applies partial action by requiring only a local quorum at one site rather than a full quorum across all sites. This partial quorum approach is sufficient for read operations because the write operation already ensured the data was replicated to a quorum at each site. The excessive action would be querying all sites; the patent reduces this to querying just one site's local quorum.
2Productivity
If a low consistency level is used for write operations, then write speed is improved, but data consistency between read and write operations deteriorates
Solution Approach 1:
The patent segments the consistency requirement by site. Each site maintains its own quorum independently, and write operations need only achieve a quorum at the coordinating site (not all sites). This segmentation allows fast writes (by not waiting for all sites) while ensuring consistency (by guaranteeing the data exists in at least one site's quorum, which can then serve read operations).
Solution Approach 2:
The patent applies partial action by requiring only a local quorum at the coordinating site for write operations, rather than requiring quorums at all sites. This partial quorum approach is sufficient because the replicated data will be available in at least one site's quorum for subsequent reads. The excessive action would be waiting for all sites to acknowledge; the patent reduces this to waiting for just the coordinating site's quorum.
3Reliability
If an each-quorum consistency level is used across all sites, then data consistency is improved, but system complexity and resource consumption increase
Solution Approach 1:
The patent segments the quorum requirement by allowing different quorum levels for different operation types and sites. Read operations use any-local-quorum (simpler), while write operations use each-quorum (stricter). This segmentation of consistency requirements reduces overall system complexity compared to uniformly applying each-quorum to all operations, while still maintaining data consistency through the write operation's stricter requirement.
4Reliability
If all replica nodes are queried for read operations, then data availability is guaranteed, but resource consumption and time increase significantly
Solution Approach 1:
The patent segments the read operation scope from all replica nodes to just one local quorum. Instead of querying all replicas across all sites, the system queries only the local quorum at a single site. This segmentation dramatically reduces resource consumption (fewer nodes queried) while maintaining data availability (the data is guaranteed to exist in at least one site's quorum due to the write operation's consistency requirements).
Solution Approach 2:
The patent applies partial action by querying only a local quorum (partial set of replicas) rather than all replica nodes. This partial querying is sufficient because the write operation ensured the data was replicated to a quorum at each site. The excessive action would be querying all replicas; the patent reduces this to querying just one local quorum, significantly reducing resource consumption.
Data Source
AI summary
A storage application is disclosed herein that employs a method for managing read and write requests associated with a database distributed across multiple replicas in multiple data centers. The storage application may be implemented on one or more computing devices in the context of a data center, cloud environment, or the like. The method includes the storage application receiving a client request to perform an operation with respect to an object at an any-local-quorum consistency level. In response to client request, the storage application iteratively generates and submits coordinator requests to the multiple data centers at a local-quorum consistency level, until one of the multiple data centers successfully replies to one of the coordinator requests with a result. The storage application then replies to the client request based on the result provided by the one of the multiple data centers.