Block Storage provides network-attached block devices for cloud hosts — from cost-efficient mechanical disks to enterprise NVMe SSDs with RDMA delivering 1.2 million IOPS and sub-0.2ms latency. Attach, expand, snapshot, and migrate disks without rebooting.
RSSD Cloud Disk uses NVMe SSD media with RDMA network transmission — delivering local-disk performance over the network. These are not theoretical peaks: they are achievable at production capacity.
Block Storage gives cloud hosts a reliable, expandable block device with an independent lifecycle. Detach from one host and reattach to another. Expand without rebooting. Snapshot at any point.
Choose by your I/O pattern and latency requirement. All types share the same independent lifecycle, online expansion, snapshot, and clone features — only the underlying media and network transmission differ.
Mechanical disk storage for sequential workloads where cost per GB matters more than IOPS. Ideal for log storage, backups, and large infrequently-accessed datasets.
NVMe SSD storage for general production workloads. Suitable for databases, boot disks, and medium-intensity I/O applications requiring consistent sub-3ms response.
NVMe SSD with RDMA network bypasses host CPU for IO — delivering latency comparable to local-attached storage. For mission-critical databases and latency-sensitive applications.
Four capabilities that apply to every cloud disk type — regardless of whether you choose Regular, SSD, or RSSD. The performance tier changes; these features stay constant.
The key advantage of cloud block storage over a physical hard drive is its independence. A cloud disk follows your workflow — not the hardware's physical constraints.
Unlike a physical hard drive soldered into a server, a cloud disk exists as a network-attached block device — independent of any host. This decoupling is the architectural property that makes all other cloud disk features possible.
Cloud disk type should be driven by I/O pattern. Sequential writes → Regular is often enough. Random reads with microsecond latency requirements → RSSD is the only answer.
MySQL, PostgreSQL, MongoDB, Redis — all benefit from sub-millisecond storage. RSSD's 1.2M IOPS and 0.1ms latency eliminates storage as a bottleneck even at extreme qps.
Operating system, application binaries, and runtime libraries. SSD provides fast boot times, quick application starts, and responsive OS I/O — at a lower cost than RSSD for this pattern.
Sequential writes for application logs, database backups, and cold data don't need low latency. Regular Cloud Disk delivers 100 MB/s sequential throughput at the lowest cost per GB — ideal for high-volume log accumulation.
GPU compute jobs read model checkpoints and training batches at high throughput. RSSD's 4.8 GB/s throughput means large checkpoints load in seconds, not minutes — keeping GPU utilisation high between iterations.
From cost-efficient mechanical storage to 1.2M IOPS enterprise NVMe with RDMA. Independent lifecycle, online expansion, snapshots, Data Ark — all without rebooting.