Redundant Rack Storage Network for Low-Latency Data Durability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing block-based storage systems face challenges in maintaining data durability and low latency due to single-point failures, such as server node failures or common control plane failures, which can lead to significant storage capacity unavailability and high recovery latencies across multiple locations.
Innovation Solution
A data storage system architecture with multiple head nodes and data storage sleds, where data is replicated across multiple mass storage devices, allowing for low latency access and high durability by designating primary and secondary head nodes and using erasure encoding across multiple sleds, enabling autonomous operation without relying on a zonal control plane for read/write requests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is stored across multiple devices in multiple locations to increase durability, then data durability is improved, but access latency increases and service availability decreases upon failure
Solution Approach 1:
The system segments storage into hierarchical tiers: hot data resides in local head node storage for low-latency access, while cold data is segmented across multiple data storage sleds. This segmentation allows frequently accessed data to be served locally without the latency penalty of distributed access, while still maintaining durability through replication and erasure coding across sleds for non-frequently accessed data.
Solution Approach 2:
The patent introduces a new dimension of spatial organization by col locating head nodes and data storage sleds within the same rack. This dimensional reorganization reduces physical distance and network hops between storage and compute resources, thereby reducing access latency while maintaining the benefits of distributed storage across multiple sleds for durability.
2Ease of operation
If a common control plane is used to manage data in multiple locations, then system management is simplified, but a failure of control plane components impacts large quantities of storage capacity and reduces service availability
Solution Approach 1:
The control functionality is segmented into distributed head nodes, each capable of independently managing its local storage and participating in the erasure coding scheme. This segmentation eliminates the single point of failure in centralized control planes, as each head node can continue to service requests and maintain data even if other head nodes or control components fail.
Solution Approach 2:
Head nodes are designed to autonomously perform data management functions including encoding data into erasure coded segments, managing local storage, and participating in data recovery operations without requiring centralized control plane coordination for every operation. This self-service capability ensures continuous operation during control plane failures.
3Adaptability or versatility
If extensive networks are used to move data between multiple locations, then data distribution capability is improved, but data recovery latency increases due to data being located across multiple locations
Solution Approach 1:
The network architecture is segmented into hierarchical levels: fast intra-rack networks connect head nodes to local sleds for rapid data access, while inter-rack networks provide extended distribution capability. This segmentation allows the system to utilize high-speed local networks for frequent operations and slower long-distance networks only when necessary, thereby reducing recovery latency while maintaining broad data distribution capability.
Solution Approach 2:
The patent adds a spatial dimension to network architecture by col locating storage sleds with head nodes in the same rack, creating a hierarchical network structure. This dimensional organization places frequently accessed data physically close to processing nodes, reducing network distance and latency for data recovery operations while still allowing data to be distributed across multiple racks when needed for durability.
Data Source
AI summary
A data storage system includes a rack, multiple head nodes, multiple data storage sleds, and at least two networking devices. The at least two network devices are configured to implement at least two redundant networks within the data storage system. Also, each of the head nodes is assigned at least two network addresses for communication with the data storage sleds of the data storage system via the at least two networking devices. The data storage sleds each include multiple mass storage devices and a sled controller that is configured to couple with the at least two network switches. IO In some embodiments, the data storage system further includes redundant power systems within a rack in which the head nodes, the data storage sleds, and the at least two networking devices are mounted.