Storage analyst Fred Moore argues that enterprise storage lacks a technology combining the access speed of disk with the low operating cost and energy use of tape. In a report titled “The Missing Storage Tier,” he assesses nine approaches to long-term data retention and concludes that none currently meets the full set of requirements for an integrated archive platform.
Moore, president of Horison Information Strategies, frames the gap between primary storage, where access is typically measured in milliseconds, and archival storage, where retrieval can take minutes, hours or days. He says the traditional archive tier has split into three classes according to how often data is retrieved: Active Archive, Archive and Deep Archive.
His labels for those classes are WORM, or Write Once, Read Many; WORSe, or Write Once, Read Seldom; and WORN, or Write Once, Read Never. Their differing access patterns and data values make it difficult to serve them cost-effectively with a single existing medium. Active archive data is generally held on hard drives, and sometimes SSDs where latency or bandwidth needs exceed what HDDs can provide.
Current active-archive designs can place an SSD or HDD cache in front of tape libraries or cloud storage, with software managing movement between layers. Moore says this arrangement works but creates operational overhead when large datasets have to be recalled, copied or staged again. He proposes a single system able to serve all three archive classes while avoiding repeated data movement.
Requirements for the proposed tier
The report describes a target tier with access in seconds and random access after media is loaded. Its cost per terabyte would be lower than HDD and approach tape economics, while total cost of ownership would be low. The media should consume no energy while idle, require no forced migration cycles and last for decades or longer.
Moore also identifies removable media, air-gapping, WORM immutability and physical vaulting as protections against cyber incidents and disasters. The system should scale from petabytes to exabytes and accommodate bursts of AI-related retrieval. He says SSDs and HDDs can deliver performance, while tape offers economic, energy and security advantages, but no current technology combines both sets of attributes.
The envisioned design uses removable random-access media, likely in a robotic library. Moore says an ideal medium would not need to rotate, spin or move during access; a read/write mechanism could instead cover its surface with minimal energy. These characteristics remain design goals rather than capabilities he says are available in a proven product.
Nine technologies assessed
Moore scored nine suppliers or projects against 13 requirements, with each criterion weighted equally and a maximum score of 65. A five indicates a proven or inherent capability, three a credible roadmap target or partial fit, and one a requirement not met. Where public information was unavailable, the report assigned two points, making lack of evidence a penalty.
- Cerabyte: 55
- Folio Photonics: 50
- SPhotonix: 48
- HoloMem: 46
- Microsoft Project Silica: 44
- Atlas DataStorage: 42
- DOTS/Indellium: 41
- Biomemory: 41
- Piql: 40
Five criteria—energy, data movement, air gap, removability and immutability—received identical scores across the candidates, according to the report. Moore says the ranking therefore turns on factors including price, access time, data rate, random access, scalability, support for hot and cold workloads, and roadmap maturity.
The report characterizes Cerabyte as best positioned against its definition of the missing tier, while noting that the definition aligns closely with the company’s own messaging. It places glass-based approaches such as Microsoft Project Silica and SPhotonix, film or tape approaches including Piql and DOTS, and DNA storage from Atlas and Biomemory closer to deep-archive or time-capsule use. DNA is described as especially suited to an ultra-cold tier because of high read and write costs and long latency. HoloMem is identified as a potential near-term improvement to tape-based archiving.
Moore’s conclusion is that the need is defined, but no candidate has yet demonstrated both archival-level cost and practical access performance. The report identifies Cerabyte’s 2028–2030 rack-scale targets, Folio Photonics’ first commercial deployments and HoloMem’s stated move to mass production in 2027 as milestones to watch. Those are forward-looking targets, not evidence that the products have already met the proposed requirements.
The report file was distributed under Cerabyte’s name, and the material notes that the company markets its product as “The Missing Storage Tier.” It also says Cerabyte effectively paid for the report, a context readers should consider when evaluating its comparative rankings and conclusions.