English
FusionStor® S6000 Series
    Publish time 2017-02-22 13:28    


Product Overview

FusionStor® S6000 series is a distributed block storage product independently developed by Beijing Fusion Stack with completely independent intellectual property rights. It gathers the storage medium installed in multiple servers through Ethernet or Infiniband network to form a storage resource set with high availability, easy management, easy scaling and high performance, which can define the semantics based on the application layer. In addition, it can distribute the storage space flexibly according to needs and provide to the upper application system through iSCSI protocol NBD protocol, QEMU driver, OpenStack Cinder and many other interface protocols.

The product has a strong horizontal scalability (Scale-Out), excellent performance (high IOPS, low latency), and has extensive application value in cloud platform, virtualization, database and other scenes.

Product Characteristics

Bare metal performance

  • Advanced architecture: It adopts the advanced Kernel-bypass architecture to bypass the call of a lot of operating system layers, greatly shortening the IO route.
  • Leading technology: In the high-speed and low-latency network (Infiniband), S6000 adopts the deep performance optimization of RDMA, iSER, DPDK and other technologies.
  • Polling technology: It uses Polling technology to lock a small amount of CPU cores, so as to give play to the processing power of CPU core to the most extent and reduce the overhead generated due to scheduling between CPU cores.
  • Layering and caching: It uses SSD, NVMe and other high-performance storage media as a cache or hot data layer to enhance the overall storage performance.
  • Data localization: In virtualization and other scenarios, the user can use data localization technology to keep data copies in the nearest node from the business, reducing overhead for accessing to network.

Stability and reliability

  • Data redundancy: Provides volume-level redundancy strategy customization, which can be used to select 2-4 replica strategy according to the important level of data, or N+1, N+2, N+3 and N+4 erasure code strategies with higher space utilization.
  • Fault domain: Support the rack topology as a fault domain for partition rules to avoid the same data redundancy blocks stored in the same fault domain and avoid data unavailability due to outage of the equipment cabinet.
  • Protection domain: Servers with different performance, network access and different service life in the super-large cluster can be divided into different protection domains and the storage volumes can be stored without across domain. It can play a buffer role in frequent faults at the final stage of the service life of the hardware and can isolate the servers according to different network accesses.
  • Intelligent recovery: The system automatically detects multiple faults in the disk, network and nodes and takes the initiative to conduct a fault recovery. According to the predefined recovery QoS strategy, it can avoid the impact on business access performance while recovering.

Elastic scaling

  • Online node scaling: It uses the Scale-out horizontal scalability to dynamically add the nodes without interrupting service and expand the cluster size of 256 nodes.
  • Liner scaling of capacity: The storage space of the cluster increases with the node, showing a linear growth; the size of the storage volume is dynamically scalable.
  • Linear improvement of performance: The IO processing capability of the entire cluster can be dynamically improved with the increase of the node.

Convenient management

  • Unified management: With regard to cluster creation, node addition, space management, log management, all management and maintenance work shall be completed by a unified visualization platform, so that the management and maintenance personnel can easily maintain the cluster with hundreds of nodes, reducing management and maintenance costs.
  • Global monitoring: It provides a graphical monitoring view for the entire cluster in terms of the status of the storage nodes, disks and volumes, so that the faults can be easily and quickly discovered and located, and it also supports e-mail and SMS alarms.
  • Status return: It has the built-in CallHome function, so that the storage cluster can take the initiative to report the status of operations to the unified monitoring and analysis platform of FusionStack to help users predict the risks in advance that may exist in the system.

Characteristic Functions

  • Super-high performance: With NVMe as a caching or layered acceleration, each CPU core of the processor can provide 200,000 IOPS, and the single storage node can provide 1.8 million IOPS (NVME x 3, IB x 2); Erasure code: Use EC algorithm to safeguard the security of the data block through the verification mode, while saving space utilization.
  • Layering+caching: Based on high performance and low latency of SSD or NVMe, it may store the volumes in the high-performance medium as always which require high IO processing ability, while common volumes can be put in between high-performance medium and high-capacity medium according to the changes in data hotspots. In addition, it uses SSD space as the buffer memory space for reading and writing, so as to speed up reading and writing of non-hotspot layers.
  • Consistency group: Provide the consistency snapshot of multiple volumes at the same moment and provide a guarantee for security rollup of virtualization, database and other applications.
  • Built-in storage network: Small and medium-sized scenarios with nodes below 4 support the direct connection mode of node network, which does not need the private exchange for storage.
  • System status return: The cluster will actively report the health indicators to the original data analysis platform.

Technical Specifications of the Product



上一个: None
下一个: FusionStac......