Distributed Block Storage ZBS

Data Block Checksum at Hard Disk Level

Deal with silent data corruption through data checksum.

Data Protection at Node Level and Automatic Data Recovery

Data protection is performed among nodes through a multi-replica or erasure coding mechanism; when a component or a node fails, the available space is automatically used, and concurrent data recovery starts among multiple nodes, ensuring that the data redundancy always meets expectations.

Protection at Rack Level

Through rack topology configuration, data is automatically placed on different racks to prevent the cluster from being inaccessible caused by power outage or failure of a single rack, further improving the storage reliability.

Snapshot Protection

By generating a snapshot for the storage, the data can be quickly restored to the state at the time when the snapshot was taken, ensuring data security.

Cross-Site Active-Active Clustering

Along with high availability configuration at the client, the stretched active-active cluster guarantees zero RPO and near-zero RTO, providing application availability during disasters.

Intelligent Recovery Policy of Business-First

On the premise of ensuring business I/O, adaptively adjust the speed of recovery or migration according to the system load.

Agile Recovery Mechanism

During node upgrade or maintenance, the expected loss of replicas will not trigger data recovery, and the write requests during the offline period of the replicas will complete data recovery with a smaller granularity after the node is restored.

Abnormal Disk Detection and Isolation

Automatically detects and isolates unhealthy, failing or low-life disks to reduce the impact on system performance and operations and maintenance.

Network Fail-Slow Detection and Isolation

Automatically and regularly checks storage and access networks and immediately isolates abnormal nodes and NICs to reduce the impact on system performance.

Fully Distributed Architecture

The distributed architecture eliminates controller bottlenecks, and the concurrent performance increases linearly with the number of nodes.

Proprietary File System Based on Bare Metal Devices

A file system is directly built on bare devices, more suitable for accessing high-performance block storage, avoiding the overhead of the existing Linux file system.

All-Flash Support

Supports all-flash storage environments to fully meet enterprises' needs for high-performance scenarios.

Automatic Tiering of Hot and Cold Data

Cold data automatically sinks to HDD, and hot data remains in the cache layer, fully making use of the advantages of SSD hardware and further improving performance.

Volume Pinning

Prevent performance degradation caused by cache breakdown by storing storage volume data in the cache layer, ensuring a consistent high performance.

High-Performance I/O Link

When Boost mode is enabled on the cluster, the vhost protocol shares memory between Guest OS, QEMU, and ZBS to optimize I/O request processing and data transfer, improving VM performance and reducing I/O latency.

High-Performance Data Transmission

The data is exchanged between cluster nodes through the RDMA protocol, effectively increasing the cluster throughput and reducing the latency.

Powerful Expansion Capability

Starting with 3 nodes, the capacity and performance can be easily expanded online in a single storage pool.

Intelligent Data Migration

Dynamically balance the data distribution within the cluster, and quickly restore the balance of data distribution after storage capacity expansion.

Intelligent Data Provision

According to the load status of the cluster capacity, the data is intelligently distributed and dynamically adjusted according to the principles of local-first, topological security, localized provision, and capacity balance, to achieve a balance of high performance and high reliability.