Live Origin Storage for Immutable Caching and Low-Latency Reads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional live origin servers are not optimized for live streaming, leading to slow read requests due to the lack of strong read after write consistency, high latency from single cloud region operation, and inefficient deletion of small media content item segments, which can cause delayed availability and increased latency.
Innovation Solution
Implement a distributed datastore with selective consistency support for mutable and immutable data, caching immutable data for faster access, hedging read and write requests, and using multi-cloud region awareness to tolerate network jitter.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional live origin servers implement strong read after write consistency, then data accuracy is improved, but processing speed deteriorates
Solution Approach 1:
The patent segments data into mutable and immutable categories, applying different consistency strategies to each. Immutable data (media content segments) is stored with simplified consistency checks, while mutable data (manifests) receives full read-after-write consistency validation. This segmentation allows the system to maintain data accuracy for critical updates while improving processing speed for bulk media operations.
Solution Approach 2:
The system applies different quality levels of consistency control to different data types. For immutable media segments, the system uses a more lenient consistency model that prioritizes speed, while for mutable manifests, it enforces strict consistency. This local differentiation of quality control resolves the contradiction by optimizing each data type's handling according to its specific requirements.
2Device complexity
If conventional live origin servers operate in a single cloud region, then system complexity is reduced, but latency increases due to network jitter
Solution Approach 1:
The patent extends the system from a single cloud region to multiple cloud regions, adding a geographic dimension to the architecture. This multi-region deployment allows the system to bypass network jitter in any single region by routing requests to alternative regions, thereby reducing latency without fundamentally complicating the core system logic through standardized replication patterns.
Solution Approach 2:
The system performs preliminary actions by pre-replicating data across multiple cloud regions before requests arrive. This proactive data distribution ensures that when requests are made, data is already available in multiple locations, reducing latency caused by network jitter. The preliminary replication establishes a foundation that enables fast response times without adding complexity to real-time request handling.
3Quantity of substance
If conventional live origin servers delete many small media content item segments, then storage is optimized, but deletion time increases significantly
Solution Approach 1:
The patent merges individual deletion operations into batch deletion processes. Instead of deleting small media segments one by one, the system groups multiple deletion requests into batches and processes them collectively. This merging of operations maintains storage optimization by removing unnecessary segments while dramatically reducing the total deletion time through economies of scale in the deletion process.
Solution Approach 2:
The system creates copies of deletion metadata or deletion plans before executing actual data removal. This allows the system to prepare and validate deletion operations in advance, organizing multiple small segment deletions into efficient batch operations. The copying mechanism enables the system to optimize storage by coordinating deletions across multiple segments without the overhead of individual deletion processing for each segment.
4Reliability
If conventional live origin servers store multiple versions of media segments, then read after write consistency is maintained, but storage space is wasted
Solution Approach 1:
The patent extracts the version management requirement from immutable media segments and applies it only to mutable manifests. By taking out the need for version tracking from media content (which is inherently immutable once published), the system maintains consistency for critical metadata while eliminating redundant storage of multiple media segment versions. This selective extraction of version control requirements resolves the contradiction between consistency and storage efficiency.
Solution Approach 2:
Instead of storing multiple versions of media segments and managing them, the system inverts the approach by storing a single version and using content-addressable storage with hashing to ensure consistency. The identity of the media segment is determined by its content hash, so the same content always resolves to the same storage location. This inversion eliminates the need for version tracking while maintaining consistency, thereby saving storage space.
Data Source
AI summary
One embodiment of a method for storing data includes determining whether first data to be stored is mutable data that can be modified or immutable data that is not modified, in response to determining that the first data is mutable data, storing the first data in a datastore, and in response to determining that the first data is immutable data, storing the first data in the datastore and in a cache.


