Massive Face Image Recognition Project For An Artificial Intelligence Company
Applications
After More Than 20 Years Of Development, Blade Has Continuously And Deeply Grasped Market Demand And Accumulated A Large Number Of Customer Resources And Rich Service Experience.
Case Background
The Police Portrait Comprehensive Platform Adopts The Industry’s First “Cloud + Edge + Terminal” Technical Architecture, Guided By The Ai portrait Business Application Of Public Security Video Images, And Carries Out Intelligent Recognition And Analysis With Portrait As The Core, Realizing Efficient Retrieval, Accurate Recognition, Large-scale Control And Spatiotemporal Analysis Of Massive Portraits. It Is Widely Used In Various Business Scenarios Such As Smart Policing, Social Governance, And Industry Supervision, Providing Accurate, Efficient And Stable Ai Portrait Application Services. The Platform’s Support Platform Uses Software-defined Storage Technology To Provide The Platform With Safe And Reliable Storage Resources.
A Shenzhen Company Specializing In The Field Of Visual Artificial Intelligence Is Committed To Building A “Visual Intelligence Acceleration Platform” Based On Visual Chips, Deep Learning And Big Data Technologies, Providing Visual Intelligence Application Solutions And Development Platforms For Thousands Of Companies In Industries Such As Safe Cities, Smart Business, Industrial Intelligent Manufacturing, Unmanned Systems, And Robots. The Application Of Massive Face Image Recognition Has First Posed New Challenges To Data Storage And Management. The Company Urgently Needs A Stable And Secure Storage Management System To Support The Rapid Growth Of Its Business.
Case Analysis
Powerleader Distributed Storage Network Topology Diagram
Solution
Overall Objective:
It Provides Stable Bandwidth And High-speed File Query, Greatly Improves The Io Performance Of Computing Tasks, And Improves The Work Efficiency Of The Cluster.
User Needs:
1) Expandable Massive Storage Space
The Capacity Can Be Expanded To The PB Level By Adding Storage Node Devices While The Business System Is Uninterrupted, So That The High-performance Computing Cluster Can Store A Large Amount Of Computing Materials And Results.
2) High Performance
It Can Support The Infiniband Network To Provide Stable Bandwidth And High-speed File Query, Greatly Reduce The Data IO Time Of Computing Tasks, And Improve The Work Efficiency Of The Cluster. By Adding IO Nodes And Storage, The Bandwidth Can Increase Linearly To Meet The Requirements Of Large-scale Data Parallel Reading And Writing.
3) Global Data Sharing
It Provides A Single System Image, And All IO Nodes Share The Same File System View, So That All Computing Nodes Can Obtain The Required Data Instantly Without Data Migration, Which Greatly Improves The Utilization Efficiency Of The Cluster, Supports Load Balancing, And Single Point Fault Tolerance.
A 3-Node PL-XDFS Cluster Is Used To Build A Distributed File Storage System And Configure Replica Volumes. Each Server Node Is Configured With 12 18TB SATA Disks, Double Replica Protection, And An Available Capacity Of 118.8TB, Meeting Current Capacity Requirements.
Subsequent Capacity Growth Can BeDynamically Expanded By Adding Nodes, Achieving Both Capacity And Performance Improvements.
The POSIX Interface IsUsed To Achieve Storage Access Through A 10G Network, Providing A Maximum Bandwidth Of More Than 1.5gb, Meeting The High-performance And Low-latency Requirements Of Computing Node Clusters For Data Computing.
Customer Benefits
Fully Distributed Massive Storage Cluster
Fully Distributed Storage Architecture, No Independent Metadata Server Design, Eliminates Metadata Information Interaction Bottlenecks, And Improves System Scalability; Scale-out Dynamic Expansion, Fast Online Addition And Deletion Of Nodes; Global Namespace, Providing Clients With A Unified Logical View And Resource Access Portal
Flexible Storage Architecture
Storage Nodes Can Be Added Or Reduced Online; Virtualized Storage Resource Pools Can Be Changed Dynamically; Capacity Nodes Can Be Selected For Initial Expansion Or Capacity Expansion Configuration According To Needs
High Reliability And Availability
Fully Active Redundant Nodes, Which Can Take Over The Fault Within Seconds After A Node Failure; Automatic Data Replication Protection And Intelligent Self-healing Mechanism
Superior Performance
Multiple Data Distribution Mechanisms And Combinations, Support For 10G And Infiniband Networks, Optimization Of Concurrent Read And Write Performance; End-to-end Automatic Load Balancing And Multi-link Aggregation Technology;
Convenient Management
The Multi-language Web-based Graphical Management Interface Allows For Easy Configuration, Monitoring, And Management. It Supports Command Line Management And Can Be Integrated With The User's Own Operation And Maintenance Platform. It Integrates With AD And LDAP To Achieve Complex User And Permission Management. It Supports Quota Management To Achieve Unified Storage Management And Allocation.