Hot, warm, and cold storage tiers
Hot, warm and cold storage tiers classify data by access demand and service requirements to balance performance and cost.
Temperature describes a placement policy, not a physical medium. A hot tier prioritizes active workloads; warm and cold tiers trade some access capability or cost structure for cheaper capacity. Boundaries depend on the service and workload. [2]
Online and offline are separate distinctions: a block-blob service's Cold tier supports direct payload access, while its Archive tier requires rehydration. Minimum billable duration can create an early-deletion charge without preventing deletion. [3]
Evaluate retrieval steps, request frequency, bulk-recovery throughput and all relevant charges before migrating data. A tier label alone does not establish redundancy, immutability or an independent recovery copy.
For a replicated HDFS 3.3.5 file with n replicas, HOT requests n on DISK, WARM one on DISK and n-1 on ARCHIVE, and COLD n on ARCHIVE; fallback rules can affect actual placement. Changing a policy does not itself relocate existing blocks; movement requires the mover or storage-policy satisfier. Its ARCHIVE storage type is not a promise of tape or offline retrieval. [5]
Relations
- related
- Enterprise SSD, Nearline HDD, Magnetic tape storage, LTO (Linear Tape-Open), Glass storage, Hybrid flash array, Warm storage tier, Hot storage tier, Cold archive tier, Automated storage tiering, Hierarchical storage management
Advantages
- Lifecycle placement can use measured changes in demand instead of keeping every object in the same layout. [2]
- Infrequent-access classes can retain prompt reads when the application cannot wait for asynchronous restoration. [4]
Limitations
- Object size, transitions and retrieval volume can change which access class is economical. [4]
- Account type and object format can restrict available tiers and transitions. [3]
Citations
- Muralidhar et al., f4: Facebook's Warm BLOB Storage System, OSDI 2014, sections 3, 6.1 and 6.2
- Microsoft Learn: Access tiers for blob data, online tiers, archive and billing
- Amazon S3 User Guide: Understanding and managing S3 storage classes
- Apache Hadoop 3.3.5: Archival Storage, SSD & Memory
Further reading
- Information lifecycle management
- Muralidhar et al., f4: Facebook's Warm BLOB Storage System, OSDI 2014
Sources & caveats
- The synthesis is not a standardized hot/warm/cold taxonomy. f4 is an immutable-blob system described at OSDI 2014.
- The Azure and Amazon S3 documents describe their own service contracts; no prices or hidden cloud media assignments are inferred.
- Historical citation IDs c2 and c3 are retained; their presence does not reinstate the older universal medium assignments.