https://hesela.com/Editorial dates for storage analyses published by Hesela team.enhttps://hesela.com/analyses/3d-nand-retention-evidence/https://hesela.com/analyses/3d-nand-retention-evidence/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTSeparate NAND architecture, dwell, thermal annealing and read conditions before interpreting raw error rates or claims of restored SSD endurance. https://hesela.com/analyses/afr-rate-not-probability/https://hesela.com/analyses/afr-rate-not-probability/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTWhy exposure, zero-failure probability bounds, survival curves and censoring assumptions matter when interpreting drive reliability, with reproducible synthetic examples. https://hesela.com/analyses/atomic-rename-directory-durability/https://hesela.com/analyses/atomic-rename-directory-durability/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTFile contents, directory entries and commit acknowledgements have different persistence boundaries. Linux documentation and two USENIX studies explain what a rename does not guarantee. https://hesela.com/analyses/backup-verification-is-not-a-restore-test/https://hesela.com/analyses/backup-verification-is-not-a-restore-test/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTSeparate backup completion, integrity, freshness and usable recovery, with reproducible SQLite and rsync experiments and explicit limits. https://hesela.com/analyses/bloom-filters-lsm-read-cost/https://hesela.com/analyses/bloom-filters-lsm-read-cost/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTLSM lookup costs, filter retention and counting limits, with reproducible allocation, deletion and memory examples grounded in primary sources. https://hesela.com/analyses/checksums-identity-and-lost-writes/https://hesela.com/analyses/checksums-identity-and-lost-writes/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTChecksums, identity, freshness, Linux verity and disk encryption: what successful verification does and does not prove, with reproducible examples. https://hesela.com/analyses/emmc-ufs-rpmb-boundaries/https://hesela.com/analyses/emmc-ufs-rpmb-boundaries/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTHow managed flash differs from raw NAND, why eMMC and UFS are not interchangeable interfaces, and why RPMB authentication is neither encryption nor a backup. https://hesela.com/analyses/ext4-journal-replay-revocation/https://hesela.com/analyses/ext4-journal-replay-revocation/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTJBD2 replay, checksum failure paths and ambiguous log tails, with source-checked recovery boundaries and reproducible counterexamples. https://hesela.com/analyses/fsync-failure-retry-durability/https://hesela.com/analyses/fsync-failure-retry-durability/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTWhy a successful fsync retry can leave stale data: historical research, Linux close errors, POSIX portability, synchronization boundaries and reproducible descriptor-lifetime observations. https://hesela.com/analyses/io-uring-completion-ordering-durability/https://hesela.com/analyses/io-uring-completion-ordering-durability/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTio_uring write ordering, protection information, cancellation CQEs and linked timeouts: what each result proves, and what still needs synchronization. https://hesela.com/analyses/iscsi-identity-sessions-recovery/https://hesela.com/analyses/iscsi-identity-sessions-recovery/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTSeparate iSCSI names, portals, sessions, command ordering and recovery levels, with a source-checked distinction between projected and measured performance. https://hesela.com/analyses/magnetic-retention-time-and-probability/https://hesela.com/analyses/magnetic-retention-time-and-probability/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTA reproducible two-state model separates first reversal, final-state error and magnetization decay, and shows why a mean energy barrier cannot predict retention. https://hesela.com/analyses/ml-disk-failure-prediction/https://hesela.com/analyses/ml-disk-failure-prediction/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTLu et al.'s fleet study tested SMART, performance and location signals. Interpret its scores alongside explicit alert denominators, prevalence and validation limits. https://hesela.com/analyses/mmap-visibility-size-durability/https://hesela.com/analyses/mmap-visibility-size-durability/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTSeven reproducible Linux checks distinguish mapped visibility, file size, partial-page tails, SIGBUS and sync argument errors from power-loss durability. https://hesela.com/analyses/nfs-stable-writes-flush-fua/https://hesela.com/analyses/nfs-stable-writes-flush-fua/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTNFSv3 stability, COMMIT verifiers and Linux flush/FUA: distinguish durable writes from cache visibility, with bounded NFSv4 delegation comparisons. https://hesela.com/analyses/nvme-fabrics-transports-discovery/https://hesela.com/analyses/nvme-fabrics-transports-discovery/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTDistinguish NVMe/TCP, NVMe/RDMA and NVMe/FC from discovery, path selection and security, without treating a transport label as a performance guarantee. https://hesela.com/analyses/o-direct-alignment-by-kernel/https://hesela.com/analyses/o-direct-alignment-by-kernel/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTQuery O_DIRECT per-file alignment and distinguish cache hints, prefetch, NOWAIT and mincore residency from actual I/O and durable writes. https://hesela.com/analyses/phi-failure-detectors-suspicion/https://hesela.com/analyses/phi-failure-detectors-suspicion/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTWhat a phi score measures, why Cassandra's score differs from the original model, and why suspecting a replica does not fence it. Includes reproducible synthetic examples. https://hesela.com/analyses/protection-information-escape-tags/https://hesela.com/analyses/protection-information-escape-tags/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTA pinned Linux verifier test separates PI escape tags, guard checks and 32-bit reference aliases. Sixteen cases show why accepted does not always mean checked. https://hesela.com/analyses/rocksdb-wal-flush-manifest-boundaries/https://hesela.com/analyses/rocksdb-wal-flush-manifest-boundaries/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTRocksDB WAL sync, memtable flush, write stalls, read snapshots, checkpoints and backups have distinct recovery and retention contracts. https://hesela.com/analyses/scrubbing-detection-not-durability/https://hesela.com/analyses/scrubbing-detection-not-durability/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTHow scrub coverage and irregular inspection intervals limit detection evidence, with exact reproducible examples of waiting-time bias. https://hesela.com/analyses/sieve-cache-eviction-traces/https://hesela.com/analyses/sieve-cache-eviction-traces/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTExplain SIEVE versus CLOCK and LRU, replay cache traces and FIFO capacity anomalies, and separate miss counts from storage performance claims. https://hesela.com/analyses/sleeping-disks-partial-reconstruction/https://hesela.com/analyses/sleeping-disks-partial-reconstruction/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTAn exhaustive check of two historical XOR codes separates serving a read without spin-up from recovering every data symbol. Includes code, all 512 availability states and explicit limits. https://hesela.com/analyses/snapshot-backup-recovery-boundaries/https://hesela.com/analyses/snapshot-backup-recovery-boundaries/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTSeparate snapshot consistency, independent recovery copies and RPO/RTO targets, with source-backed OpenZFS, PostgreSQL and VSS examples. https://hesela.com/analyses/sparse-files-preallocation-thin-capacity/https://hesela.com/analyses/sparse-files-preallocation-thin-capacity/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTFile size, allocated blocks and thin-pool headroom are different measurements. Learn what preallocation, hole punching and zero-filled reads actually establish. https://hesela.com/analyses/ssd-wear-out-failure-signatures/https://hesela.com/analyses/ssd-wear-out-failure-signatures/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTPercentage Used estimates consumed endurance; it is not a failure probability. Available Spare measures a different quantity, and a read-only warning is not a promise of continued data access. Historical SATA failures and Google's field study illustrate these limits, but neither calibrates modern NVMe warning coverage. Keep device identity, log identifier and counter scope with every observation. https://hesela.com/analyses/storage-tail-slowdowns-raid-fanout/https://hesela.com/analyses/storage-tail-slowdowns-raid-fanout/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTSlow-drive hours and request percentiles count different events. Source review and exact examples distinguish fan-out, hedged reads, cancellation work and replica dependence. https://hesela.com/analyses/tmpfile-publication-rename-contracts/https://hesela.com/analyses/tmpfile-publication-rename-contracts/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTA native Linux observation separates O_TMPFILE staging, linkat publication, no-replace rename and atomic exchange from crash durability. https://hesela.com/analyses/wal-checkpoints-torn-pages/https://hesela.com/analyses/wal-checkpoints-torn-pages/#updated-2026-10-17Sat, 17 Oct 2026 00:00:00 GMTWAL recovery boundaries: PostgreSQL hint-bit writes, page-image compression and statistics traps, plus InnoDB doublewrite and SQLite commit semantics. https://hesela.com/analyses/b-epsilon-trees-buffered-updates/https://hesela.com/analyses/b-epsilon-trees-buffered-updates/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTHow buffered tree updates remain query-visible, why blind upserts differ from read-modify-write, and why equal flush counts can hide different completion times. https://hesela.com/analyses/backblaze-age-missing-data-audit/https://hesela.com/analyses/backblaze-age-missing-data-audit/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTA full Q4 2025 raw-data audit finds 18 TB drives in a mislabeled capacity group and shows why excluding 106 missing-age drive-days also excludes 19 failures. https://hesela.com/analyses/content-defined-chunking-boundaries/https://hesela.com/analyses/content-defined-chunking-boundaries/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTReproducible FastCDC examples show how file framing, parameter changes and terminal chunks alter measured reuse, even when file contents remain unchanged. https://hesela.com/analyses/crash-testing-oracles-and-bounds/https://hesela.com/analyses/crash-testing-oracles-and-bounds/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTWhat ALICE, CrashMonkey/ACE, SQLite and Ananke actually test: fault boundaries, separate recovery outcomes and an executable correctness-oracle example. https://hesela.com/analyses/datacenter-hardware-failure-tickets/https://hesela.com/analyses/datacenter-hardware-failure-tickets/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTA hardware ticket, a failed device and a service outage are different counting units. Review the reported fleet figures, separate response time from repair time, and reproduce why a 120-second failure-burst rule need not limit a burst to 120 seconds. https://hesela.com/analyses/file-clones-deferred-copy-cost/https://hesela.com/analyses/file-clones-deferred-copy-cost/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTSeparate reflinks, hard links and in-kernel copies; trace deferred costs, Btrfs maintenance, shared-space accounting and FAST 2020's research limits. https://hesela.com/analyses/hdd-adjacent-track-interference-refresh/https://hesela.com/analyses/hdd-adjacent-track-interference-refresh/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTRepeated HDD writes can trigger protective refreshes even with CMR. Separate write-induced ATI, read-channel interference and SMR band rewrites before diagnosing latency. https://hesela.com/analyses/hdd-annualized-workload-rate/https://hesela.com/analyses/hdd-annualized-workload-rate/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTAn HDD workload rating is not SSD TBW or a failure countdown. Power-on hours, read traffic and maintenance scans change what the annualized number means. https://hesela.com/analyses/hdd-write-verification-iraw/https://hesela.com/analyses/hdd-write-verification-iraw/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTRead-after-write verification, IRAW and SCSI VERIFY answer different questions from flush or FUA. Learn what a successful check does and does not establish. https://hesela.com/analyses/lmdb-readers-page-reuse-map-full/https://hesela.com/analyses/lmdb-readers-page-reuse-map-full/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTA controlled LMDB experiment separates reader lifetime, reusable pages and map capacity: why a live reader can stop writes without holding the writer mutex. https://hesela.com/analyses/nfs-silly-rename-open-unlink/https://hesela.com/analyses/nfs-silly-rename-open-unlink/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTSilly rename, open-file preservation and a misleading Linux comment: trace NFS deletion from the 1985 paper to a pinned implementation and RFC erratum. https://hesela.com/analyses/nvme-multipath-ana-path-selection/https://hesela.com/analyses/nvme-multipath-ana-path-selection/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTSeparate NVMe namespace identity from ANA access states: NSID scope, identifier lifetime, Linux path policies and controller-level queue accounting. https://hesela.com/analyses/object-etags-checksums-conditional-writes/https://hesela.com/analyses/object-etags-checksums-conditional-writes/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTS3 ETags, full-object and composite checksums, version IDs and write preconditions: what each validates, and how to avoid checking download metadata instead of bytes. https://hesela.com/analyses/raidz-expansion-parity-capacity/https://hesela.com/analyses/raidz-expansion-parity-capacity/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTReflow spreads old blocks across more disks without changing their data-to-parity ratio. A reproducible model separates retained parity overhead, reported space and snapshot-aware rewriting. https://hesela.com/analyses/replica-repair-coverage/https://hesela.com/analyses/replica-repair-coverage/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTReplica repair coverage, quorum overlap and set reconciliation: why matching keys, successful responses and compact summaries do not prove identical durable state. https://hesela.com/analyses/replication-acknowledgements-durability-lag/https://hesela.com/analyses/replication-acknowledgements-durability-lag/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTCompare replication acknowledgements, WAL retention and logical failover readiness; distinguish durable writes, replay and missing or unusable slots. https://hesela.com/analyses/sqlite-wal-checkpoint-starvation/https://hesela.com/analyses/sqlite-wal-checkpoint-starvation/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTMeasured SQLite WAL checkpoints and stale-snapshot retries: why restarting a transaction can still overwrite an update when the application reuses old values. https://hesela.com/analyses/ssd-correlated-failures-nodes-racks/https://hesela.com/analyses/ssd-correlated-failures-nodes-racks/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTHan et al. (FAST 2021) report that 12.9% of failures were intra-node and 18.3% intra-rack at a 30-minute grouping threshold. These are shares of recorded failures, not percentages of all installed drives. Reading the analysis code also separates a group-size statistic from a prediction of the next drive failure. https://hesela.com/analyses/ssd-mapping-cache-hmb/https://hesela.com/analyses/ssd-mapping-cache-hmb/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTSeparate SSD maps from data buffers, explain cache admission and HMB, and reproduce memory, scan and request-versus-byte hit examples. https://hesela.com/analyses/ssd-power-on-hours-firmware-failures/https://hesela.com/analyses/ssd-power-on-hours-firmware-failures/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTA healthy wear indicator does not rule out a firmware-hour defect. Vendor advisories describe separate 32,768-, 40,000- and 28,224-hour issues with different affected models and recovery outcomes. Match the exact drive and firmware, then check fresh runtime evidence. https://hesela.com/analyses/ssd-thermal-throttling-failures/https://hesela.com/analyses/ssd-thermal-throttling-failures/#updated-2026-10-16Fri, 16 Oct 2026 00:00:00 GMTA hot SSD reading, a thermal-management event and a drive failure are different observations. Historical fleet studies show platform-dependent temperature associations, not a universal failure threshold. NVMe exposes distinct temperature and action counters; their units, capabilities and specification revision determine what a reading can establish.