Overview of FusionStorage Distributed Block Storage Architecture and Features
FusionStorage is a distributed block storage solution for x86 servers that separates compute and storage, uses DHT‑based key‑value architecture, provides strong consistency and read‑repair mechanisms, and supports various cache media and advanced storage functions such as thin provisioning and synchronous replication.
FusionStorage is a distributed block storage software that can be deployed on x86 servers, leveraging local HDDs, SSDs and other media to form a large storage resource pool and offering a standard iSCSI block storage interface to applications and virtual machines. It supports major server platforms including Huawei, HP DL380S, Dell, and IBM.
The architecture resembles Nutanix and supports multiple virtualization platforms such as FusionSphere, VMware, and Xen, but differs in that FusionStorage only supports SAN (no NAS or object storage) and is built on DHT technology for a key‑value store rather than a DFS.
FusionStorage Architecture
A key characteristic is the separation of storage and compute; unlike vSAN or Nutanix where they are co‑located, FusionStorage allows independent deployment, giving customers flexibility to match compute and storage requirements per application.
In a VMware environment, the compute and storage modules are separated: a CVM (control VM) provides storage services, and VMs access the storage resources via iSCSI initiators.
When storage and compute are fused, VBS and OSD are deployed together to provide storage services to local VMs.
During VMware compute‑storage fusion, host RAID disks can be passed through via PCIe to the CVM, allowing the CVM to manage storage for VMs; the ESXi layer and the CVM also reside on these disks.
FusionStorage Technical Features
Cluster Management : FusionStorage uses a cluster management approach; a failed node or disk is automatically isolated without affecting overall service.
Strong Consistency Replication Protocol : When an application writes data, all backend replicas become consistent; the write is considered successful only after the local write and acknowledgments from remote replicas, ensuring read‑after‑write consistency.
Read Repair Mechanism : If a read fails due to a disk sector error, FusionStorage reads from another replica and rewrites the faulty replica, maintaining the replica count during large‑scale linear expansion.
FusionStorage’s DHT distributes all metadata across storage nodes, eliminating cross‑node metadata access and bottlenecks. It employs data chunking and DHT‑hash routing to spread volume data evenly across a large resource pool.
Cache Technology : Supports five cache media (NVDIMM, SSD card, SSD disk, memory cache, no cache). NVDIMM and SSD provide power‑loss protection; disabling cache forces every I/O to be write‑through to disk, reducing performance.
Additional advanced storage functions include thin provisioning, synchronous replication for disaster recovery, SCSI snapshots, snapshot backup, linked clones, and inter‑node connectivity via InfiniBand, 10GE/GE.
Architects' Tech Alliance
Sharing project experiences, insights into cutting-edge architectures, focusing on cloud computing, microservices, big data, hyper-convergence, storage, data protection, artificial intelligence, industry practices and solutions.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.