Flowing Data with COS on Demand
Data Accelerator GooseFSx accelerates Cloud Object Storage (COS). As the hot tier for COS, GooseFSx stores hot data while COS provides persistent storage for full data. GooseFSx and COS exchange data on demand. GooseFSx directly loads data from COS to enable hosts to quickly access cached data in GooseFSx, while avoiding repeated data loading from COS by different hosts, reducing COS request costs. GooseFSx sinks computed results into COS for persistent, low-cost storage. GooseFSx can be released when usage is completed, with data persistently stored via COS. GooseFSx can be pulled up as needed to load data from COS and accelerate COS access.
GooseFSx exchanges specified data with COS on demand, sinking data from an entire GooseFSx directory or a certain subdirectory into COS, or loading data from an entire COS bucket or a certain prefix into GooseFSx.
GooseFSx is deeply integrated with COS, automatically loading data from COS or automatically settling data into COS. Data flow is fully managed—you do not need to deploy data migration software for data transfer, no need to intervene in the execution process of data flow, and no need to worry about configuration, tuning, management, or operation and maintenance.
GooseFSx loads data from COS, escapes objects in the COS bucket into files for hosts to access as if accessing a local file system. GooseFSx escapes files into COS objects, settles them into COS, saves them in COS in the original data format, enabling direct access to the settled data in COS.
Ease of Use
The data accelerator GooseFSx is a fully hosted service—place order and start using it immediately, with no need to worry about deployment, configuration, tuning, management, or operation and maintenance.
GooseFSx is easy to use. With just a few clicks on the console within minutes, you can complete the process from purchase and creation to mounting for use. Mount GooseFSx as a local directory on the host and use it like a local file system.
GooseFSx is POSIX file semantics compatible, allowing your workloads to use it without any modification, just like a local file system.
High-Performance and Ultra-Low Latency
The data accelerator GooseFSx adopts a fully parallel architecture designed for high-performance workloads. All nodes are fully equivalent and access data in full parallel. Metadata and data are evenly distributed across all nodes. Performance scales linearly with node expansion, delivering hundreds of GBs of throughput per second, millions of IOPS, and sub-millisecond latency.
Seamless Integration with Computing Ecosystem, Supporting Linux and Windows
The data accelerator GooseFSx is POSIX file semantics compatible, can auto mount as a local directory on the host, and seamlessly integrates into the computing ecosystem. It simultaneously supports Linux and Windows hosts, addressing issues with many applications, such as the DAE tool, which only supports the Windows platform.
GooseFSx integrates with Tencent Cloud High Performance Computing (THPC) to enable rapid deployment of high-performance computing environments. Leveraging GooseFSx's high-performance and ultra-low latency, it accelerates computational performance and shortens calculation time, enhancing overall efficiency.
Elastic and Efficient
Storage/computing separation. The data accelerator GooseFSx is separated from computing, independently expanding and upgrading, auto scaling, independently pulling up and terminating as needed, and flexibly scheduling.
Hot and cold data layering; cold data is persisted to COS, and hot data is cached in GooseFSx. COS offers unlimited capacity to meet the long-term, reliable, and low-cost large-capacity storage needs for massive data. GooseFSx supports PB-level scalability, hundreds of GB per second throughput, millions of IOPS, and sub-millisecond latency to meet high-performance access requirements.
GooseFSx and COS are decoupled from each other, scaling elastically respectively. GooseFSx and COS are deeply integrated to achieve data flow on demand.
GooseFSx uses hourly billing and can be released at any time after usage is completed. It persists the calculation results generated on GooseFSx to COS. GooseFSx can be pulled up at any time, loads data from COS, directly accesses the cached data in GooseFSx, and accelerates access to COS.