Expected Recovery Time in DNA-based Distributed Storage Systems @SNIAVideo
Expected Recovery Time in DNA-based Distributed Storage Systems  @SNIAVideo
Uploaded August 2026 | Updated November 2026, 6 hours ago
DNA-based data storage has emerged as a promising technology for long-term archival storage due to its exceptional density and durability. While most prior work focuses on storing data within a single container, practical large-scale systems will likely distribute data across multiple containers to improve robustness and availability.
In this work, we initiate the study of DNA-based distributed storage systems, in which information is encoded across several DNA storage containers so that the contents of a failed container can be reconstructed from the remaining ones. This setting is reminiscent of classical distributed storage systems, where encoded data are distributed across many servers and a failed server is repaired by downloading information from the surviving ones. Unlike servers, however, DNA storage containers are fundamentally constrained: no computation can be performed within a container, and each read operation returns the content of a uniformly random strand.
We formalize a model with M containers, each storing n distinct DNA strands. The encoded data are represented as an n × M matrix whose columns correspond to the containers and whose entries are the strands. To enable recovery from container failures, redundancy is introduced across the columns. When a container fails, a recovery process is initiated in which, at each time step, one randomly sampled strand is read from each surviving container. Our goal is to determine the expected recovery time required to reconstruct the lost column. We analyze two coding constructions:

• Scalar MDS Codes: Each row of the matrix forms a codeword of a Maximum Distance Separable (MDS) code.
• MDS Array Codes: A more general construction that leverages ideas from regenerating codes to improve recovery efficiency.

For scalar MDS codes, we derive precise asymptotic expressions for the expected recovery time, both for an arbitrary number of redundancy containers and for the case of a single redundancy container when l copies of each strand must be collected. For MDS array codes, we develop a general framework that yields upper bounds on the expected recovery time, and demonstrate through a concrete example that these constructions can outperform the scalar MDS scheme.
The questions studied here can be viewed as generalizations of the classical Coupon Collector’s prob- lem involving multiple parallel collectors. To the best of our knowledge, these variants have not been previously studied and may be of independent interest.

Presented by
Adi Levy, Technion

This is a presentation from the 2026 Storage and Computing with DNA Conference.
· Learn More about the SNIA DNA Data Storage Alliance: snia.org/groups/snia-dna-technology-affiliate
· SNIA Educational Library: snia.org/library
· X: twitter.com/SNIA
· LinkedIn: linkedin.com/company/snia
Expected Recovery Time in DNA-based Distributed Storage SystemsSNIA Swordfish Demo FMS 2026Binary Digit Trainer: Tech in a NutshellSNIA SFF Community - 2026 SNIA PreviewSNIA SDC: StorageAI 2026 - Data Storage Innovations for Scalable AI InfrastructureSNIA Swordfish TWG -2026 SNIA PreviewSNIA Computational Storage TWG - 2026 SNIA PreviewSNIA SDC 2025  - Towards Building a Flexible, Efficient & Resilient Training w/ Adaptive CheckpointAn Adaptable DNA Codec for In Memory ComputeSNIA SDC 2025  - Assessing AI Storage Communication Performance at ScaleSNIA SDC 2025  -Driving Sustainability in Datacenters -Standardizing Carbon Assessment MethodologiesSNIA SDC 2025  - Why does NVMe Need to Evolve for Efficient Storage Access from GPUs?
SNIAVideo |

Expected Recovery Time in DNA-based Distributed Storage Systems

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER