Storage Sled Write Fencing for Durable Low-Latency Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Block-based storage systems face challenges in maintaining data durability and low latency due to single-point failures in server nodes or control planes, leading to significant storage capacity unavailability and high recovery latencies across multiple locations.
Innovation Solution
A data storage system with a rack-mounted configuration of head nodes and data storage sleds, where data is replicated across multiple sleds and nodes, allowing for independent operation without relying on a zonal control plane, and utilizing redundant networks and power systems to ensure high reliability and durability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is stored across multiple devices in multiple locations using a common control plane, then data durability is improved, but system complexity and recovery latency increase
Solution Approach 1:
The system divides storage resources into discrete storage units that can be independently managed and allocated. Each storage unit operates autonomously without requiring coordination through a common control plane, eliminating the complexity of managing distributed data across multiple locations while maintaining data durability through replication across segmented storage units.
Solution Approach 2:
Storage units are designed to be self-sufficient and self-managing, capable of independently handling data operations without relying on external control plane coordination. This self-service capability reduces system complexity by eliminating the need for centralized control while maintaining reliability through autonomous operation of replicated data copies.
2Ease of operation
If a server node services multiple storage nodes, then ease of operation is improved, but reliability deteriorates due to single-point failure
Solution Approach 1:
The system segments the traditional server-node hierarchy into independent storage units, where each unit can operate autonomously. This segmentation eliminates the single-point failure risk associated with centralized server nodes while maintaining ease of operation through standardized interfaces and automated management of the segmented units.
Solution Approach 2:
The system implements dynamic failover capabilities where storage units can automatically assume roles and responsibilities of failed units without requiring manual intervention or centralized coordination. This dynamic reconfiguration maintains both reliability and ease of operation by automatically redistributing workloads when failures occur.
3Reliability
If data is located across multiple geographic locations, then durability is improved, but access speed and recovery latency worsen
Solution Approach 1:
The system implements local quality by placing complete copies of data locally within each storage unit, eliminating the need to access remote geographic locations for data retrieval. Each storage unit maintains full data availability locally while still providing durability through replication, thereby achieving both fast access speeds and high durability without the latency penalties of geographic distribution.
Data Source
AI summary
A data storage system includes multiple head nodes and data storage sleds. The data storage sleds include multiple mass storage devices and a sled controller. Respective ones of the head nodes are configured to obtain credentials for accessing particular portions of the mass storage devices of the data storage sleds. A sled controller of a data storage sled determines whether a head node attempting to perform a write on a mass storage device of a data storage sled that includes the sled controller is presenting with the write request a valid credential for accessing the mass storage devices of the data storage sled. If the credentials are valid, the sled controller causes the write to be performed and if the credentials are invalid, the sled controller returns a message to the head node indicating that it has been fenced off from the mass storage device.