Disaster Recovery Technologies: SDS, Ceph RBD Mirror, Containers, Hyper‑Converged Infrastructure, Cloud & Edge Computing, and Blockchain

Disaster‑recovery (DR) is not an isolated technology; it encompasses backup, replication, imaging, virtualization, open‑source architectures, and hyper‑converged solutions, forming complex systems that aim for a "no‑perceived‑failure" experience for users.

The article references the "China Disaster‑Recovery Industry White Paper" and provides a download link for the original document.

DR and Software‑Defined Storage (SDS)

SDS abstracts, pools, and automates storage resources. Using Ceph as an example, its strong consistency model requires all replicas to acknowledge a write, which degrades performance for geographically dispersed replicas. The RBD Mirror feature, introduced in 2015, addresses asynchronous cross‑cluster backup, operating similarly to MySQL master‑slave replication.

The Ceph journal mechanism (specific to Ceph RBD) records writes before they are flushed to the image, providing a distributed log system with replay, notification, and space reclamation capabilities.

Ceph RBD Mirror Asynchronous Backup Steps

I/O is written to the primary cluster's Image Journal.

After the journal write succeeds, RBD returns a response to the client.

The mirror process on the backup cluster detects journal updates, reads the data from the primary journal, and writes it to the backup cluster.

Once the backup cluster writes successfully, it updates metadata in the primary journal to mark the I/O as synchronized.

The primary cluster periodically cleans up journal entries that have been replicated.

Advantages include reduced write latency for remote replicas and decreased data loss risk from power failures.

While RBD Mirror continues to evolve, a complete DR solution also requires data restoration and high‑availability components, which SDS alone cannot provide; integration with databases, virtualization, and other layers is necessary.

DR and Containers

Open‑source trends such as Docker and OpenStack influence DR development. OpenStack, widely adopted (≈46% of production deployments), enables flexible switching between heterogeneous data‑center equipment and supports multi‑industry use cases.

Container‑based DR aims for simpler, open, and reusable solutions, with Docker and OpenVZ already supporting high‑availability scenarios.

DR and Hyper‑Converged Infrastructure (HCI)

HCI combines compute, networking, storage, and virtualization in a single appliance, adding caching, deduplication, compression, backup, and snapshot capabilities, and scales out horizontally.

Although HCI offers modular expansion, it can introduce single‑point‑of‑failure risks in heavily virtualized environments, making robust backup essential.

DR and Cloud Computing

Cloud computing provides on‑demand, pay‑as‑you‑go resources, reducing IT spend and enabling elastic scaling. However, reliance on external cloud services introduces security concerns such as brute‑force attacks, web‑intrusion attempts, and DDoS incidents.

Cloud DR (cloud‑based backup and recovery) mitigates these risks by distributing data across multiple locations, but bandwidth and latency constraints affect large‑scale data replication.

DR and Edge Computing

Edge computing addresses massive local data‑processing needs when network bandwidth or latency to central clouds is insufficient. By partitioning and encrypting data locally, edge nodes can store backup fragments, ensuring secure, distributed protection.

DR and Blockchain

Blockchain offers decentralized, immutable, and encrypted data storage, aligning well with backup requirements such as data security, auditability, and multi‑copy redundancy. Solutions like Storj and Sia demonstrate cost‑effective, distributed storage.

A typical blockchain‑based backup workflow includes setting a backup directory, encrypting and sending files with metadata, distributing encrypted chunks to storage nodes, and reconstructing data on demand.

The article concludes with references to the original white‑paper and a list of related reading materials.