OBS
Object Storage Service
Secure, stable, and infinitely scalable object storage that helps you save money and monetize data
Key Highlights
Proven Expertise & Capabilities
30+ Years
99.9999999999% (12 nines) of data durability.
Strong Security
Fine-grained access control, encryption, audit, and others.
High Efficiency
Tens of millions of TPS and 2.4 Gbit/s single-stream uploads and downloads.
Ease of Use
OBS Console and command line tools available for managing massive amounts of objects.
Why Us
Why Huawei
Cloud OBS?
- 99.9999999999% (12 nines) of data durability with a multi-redundant architecture
- Service Level Agreement (SLA) of up to 99.995%/li>
- Tens of millions of TPS and responses within milliseconds
- 2.4 Gbit/s single-stream uploads and downloads powered by aggregation of small objects and multi-thread parallel processing
- Fully distributed metadata management, unified management of hundreds of billions of objects in a single bucket, and efficient access to massive files
- Fine-grained access control for accounts, buckets, objects, and VPCs
- YHigh data security and reliability ensured by encryption in storage or during transfer and integrity check
- Compliance ensured by data encryption and log audits
- OBS Console, command line tools, and SDKs in over 10 languages provided for managing massive objects
- Scheduled uploads, lifecycle management, and file sharing for diverse needs
- A solution that integrates storage, processing, and governance, which is based on a combination of OBS, CDN, and MRS
- Object and POSIX protocols, and capabilities such as rename, append upload, and modify
Application Scenarios
Object Storage for Any Scenario
Scenario
An AI training and inference solution based on OBS and SFS Turbo is provided to enhance storage read/write and accelerate AI training. This further reduces GPU/NPU wait times and improves resource utilization.
Advantages
- Fast training with high-performance storage
- With the high-performance storage of SFS Turbo, GPUs/NPUs do not sit idle waiting for storage I/O operations to complete.
- Terabytes of checkpoint file data in large models can be saved and loaded in seconds.
- Short wait with asynchronous import and export
- Before training, data is pre-fetched from OBS to SFS Turbo.
- In training, checkpoint data written to SFS Turbo is asynchronously exported to OBS.
- Low costs with automated management of hot and cold data
- SFS Turbo automatically moves cold data to OBS to free up extra high-performance storage capacity.
- SFS Turbo automatically loads cold data from OBS to improve access performance.
Big Data Analytics
An AI training and inference solution based on OBS and SFS Turbo is provided to enhance storage read/write and accelerate AI training. This further reduces GPU/NPU wait times and improves resource utilization.
Advantages
- Fast training with high-performance storage
- With the high-performance storage of SFS Turbo, GPUs/NPUs do not sit idle waiting for storage I/O operations to complete.
- Terabytes of checkpoint file data in large models can be saved and loaded in seconds.
- Short wait with asynchronous import and export
- Before training, data is pre-fetched from OBS to SFS Turbo.
- In training, checkpoint data written to SFS Turbo is asynchronously exported to OBS.
- Low costs with automated management of hot and cold data
- SFS Turbo automatically moves cold data to OBS to free up extra high-performance storage capacity.
- SFS Turbo automatically loads cold data from OBS to improve access performance.
Static Website Hosting
An AI training and inference solution based on OBS and SFS Turbo is provided to enhance storage read/write and accelerate AI training. This further reduces GPU/NPU wait times and improves resource utilization.
Advantages
- Fast training with high-performance storage
- With the high-performance storage of SFS Turbo, GPUs/NPUs do not sit idle waiting for storage I/O operations to complete.
- Terabytes of checkpoint file data in large models can be saved and loaded in seconds.
- Short wait with asynchronous import and export
- Before training, data is pre-fetched from OBS to SFS Turbo.
- In training, checkpoint data written to SFS Turbo is asynchronously exported to OBS.
- Low costs with automated management of hot and cold data
- SFS Turbo automatically moves cold data to OBS to free up extra high-performance storage capacity.
- SFS Turbo automatically loads cold data from OBS to improve access performance.
Video on Demand
An AI training and inference solution based on OBS and SFS Turbo is provided to enhance storage read/write and accelerate AI training. This further reduces GPU/NPU wait times and improves resource utilization.
Advantages
- Fast training with high-performance storage
- With the high-performance storage of SFS Turbo, GPUs/NPUs do not sit idle waiting for storage I/O operations to complete.
- Terabytes of checkpoint file data in large models can be saved and loaded in seconds.
- Short wait with asynchronous import and export
- Before training, data is pre-fetched from OBS to SFS Turbo.
- In training, checkpoint data written to SFS Turbo is asynchronously exported to OBS.
- Low costs with automated management of hot and cold data
- SFS Turbo automatically moves cold data to OBS to free up extra high-performance storage capacity.
- SFS Turbo automatically loads cold data from OBS to improve access performance.
Genome Sequencing
An AI training and inference solution based on OBS and SFS Turbo is provided to enhance storage read/write and accelerate AI training. This further reduces GPU/NPU wait times and improves resource utilization.
Advantages
- Fast training with high-performance storage
- With the high-performance storage of SFS Turbo, GPUs/NPUs do not sit idle waiting for storage I/O operations to complete.
- Terabytes of checkpoint file data in large models can be saved and loaded in seconds.
- Short wait with asynchronous import and export
- Before training, data is pre-fetched from OBS to SFS Turbo.
- In training, checkpoint data written to SFS Turbo is asynchronously exported to OBS.
- Low costs with automated management of hot and cold data
- SFS Turbo automatically moves cold data to OBS to free up extra high-performance storage capacity.
- SFS Turbo automatically loads cold data from OBS to improve access performance.
Intelligent Video Surveillance
An AI training and inference solution based on OBS and SFS Turbo is provided to enhance storage read/write and accelerate AI training. This further reduces GPU/NPU wait times and improves resource utilization.
Advantages
- Fast training with high-performance storage
- With the high-performance storage of SFS Turbo, GPUs/NPUs do not sit idle waiting for storage I/O operations to complete.
- Terabytes of checkpoint file data in large models can be saved and loaded in seconds.
- Short wait with asynchronous import and export
- Before training, data is pre-fetched from OBS to SFS Turbo.
- In training, checkpoint data written to SFS Turbo is asynchronously exported to OBS.
- Low costs with automated management of hot and cold data
- SFS Turbo automatically moves cold data to OBS to free up extra high-performance storage capacity.
- SFS Turbo automatically loads cold data from OBS to improve access performance.
