Syneto Syneto
White Paper · 2026

SyFS — The Syneto Filesystem.

Copy-on-write architecture with native immutable Recovery Points, end-to-end data integrity, deduplication and compression — the storage foundation of every Syneto product.

Companies
5,000+
Partners
250+
CSAT
98%
Bytes lost
$0
SYNETO  |  WHITEPAPER-SYFS-FILESYSTEM  |  WHITE PAPER v1.0 EN-IT-ES
White Paper02
Legal imprint

About this document

Publisher

Syneto SpA

Via Cefalonia 70, Brescia 25124, Italy

VAT RO15999720 · REG J35/2971/2003

Contact: [email protected]

Phone: +39 051 095 3000

Web: https://syneto.eu

Offices

Italy (HQ) · Brescia

Romania · Timișoara

Spain · Madrid

Certifications

ISO 9001 — Quality Management

ISO/IEC 27001 — Information Security

Copyright & trademarks

© 2026 Syneto SpA. All rights reserved. "Syneto", the Syneto logo, "SynetoOS", "Syneto CENTRAL", "Hyperion", "HYPER Core", "HYPER Edge", and "HYPER Echo" are trademarks of Syneto SpA. All other trademarks are the property of their respective owners.

Document scope

This whitepaper documents the SyFS filesystem in detail: on-disk format, transaction-group commit cycle, RAIN replication groups, ARC/L2ARC caching, end-to-end checksumming and self-healing, the WORM lifecycle of Recovery Points, scrub semantics, capacity math (dedup + compression), and the operational tooling exposed to administrators.

syneto.euPage 2 of 12
Overview03
Chapter 01 · Overview

What SyFS is, what it isn't.

SyFS is a 128-bit copy-on-write filesystem derived from the OpenZFS lineage and extensively hardened for the Syneto Hybrid Cloud Ecosystem. Unlike a traditional LVM+ext4 stack, SyFS unifies volume management, RAID, snapshots, replication and checksumming into a single in-kernel subsystem.

128-bit
Address space
CoW
Write semantics
SHA-256
Per-block checksum
WORM
Recovery Point commit

Why CoW matters here

Recovery Points are free of overhead — a Recovery Point is simply a pinned set of block pointers, not a copy.
Writes never overwrite live blocks. A power loss leaves the previous transaction intact, eliminating the need for fsck.
Cloning a VM is metadata-only. New IO begins copying blocks lazily as they diverge from the parent.
Free-space management is unified with allocation — no fragmentation accumulator like ext4 inode bitmaps.
syneto.euPage 3 of 12
Architecture04
Chapter 02 · Architecture

The on-disk format.

POSIX layer
VFS hooksNFS / SMBZVOL block deviceiSCSI LIO target
DMU + ZIL
Object setsTransaction groupsIntent logAtomic groups
SPA + RAIN
Block pointers (384-bit)vdev mirrorsRAID-Z2 / Z3 paritySpare hot-swap
ARC + L2ARC
Adaptive cacheNVMe L2ARCMFU + MRU listsAsync prefetch
Physical disks
NVMe U.2SATA SSDSAS512e / 4Kn sectors
SyFS subsystem stack (top: POSIX → bottom: physical)

Block-pointer anatomy

Pointer size
384 bits
Block address
256 bits
Checksum
64 bits (Fletcher4 / SHA-256)
Birth TXG
64 bits
Why 384-bit pointers

Each pointer carries enough metadata to detect silent corruption (checksum), reason about replication (birth-txg), and locate redundant copies — without consulting a separate index. The metadata IS the address.

syneto.euPage 4 of 12
Write path05
Chapter 03 · Write path

What happens on every write.

A POSIX write enters the DMU layer, accumulates in a transaction group, and lands atomically when the group commits. The uberblock — the root pointer of the on-disk Merkle tree — is updated last. Crashes are silently safe.

write fsync commit VM / App POSIX write() DMU Open transaction ZIL (intent log) Synchronous fsync() Transaction group Buffered ~5s SPA Allocate + checksum Disks NVMe + SSD Uberblock Atomic update
Write path: POSIX → DMU → TXG → SPA → disks; uberblock updated last (atomic commit barrier).

What the uberblock guarantees

The on-disk image is always either the pre-commit state or the post-commit state — never a partial blend.
Power loss during a transaction simply replays from the ZIL on reboot. No fsck. No "metadata recovery" wait.
The last 128 uberblocks are preserved as a ring buffer — enabling rollback for advanced recovery scenarios.
syneto.euPage 5 of 12
Data integrity06
Chapter 04 · Data integrity

Self-healing in action.

mismatch match Read request POSIX read() ARC Cache lookup Mirror A Block + checksum Mirror B Redundant copy Verify sha-256 match? Self-heal Rewrite bad block Return data Validated bytes
Read path: every block is verified on the way out. Mismatches are auto-healed from the redundant copy before returning to the caller.

What this catches

Bit-rot on long-dormant data — discovered the next time the block is read, fixed automatically from the redundant copy.
Cosmic-ray and electrical-noise events that flip a single bit in a multi-terabyte file.
HBA / cable / firmware bugs that corrupt blocks between disk and RAM — checksum is computed end-to-end.
Hostile or buggy applications writing through the storage stack — they cannot bypass the SyFS write path.
syneto.euPage 6 of 12
Recovery Points07
Chapter 05 · Recovery Points

How Recovery Points work.

A Recovery Point is created by pinning the current set of block pointers and incrementing the dataset txg. The cost is constant-time, regardless of dataset size. There is no copy. There is no scan.

01

Create

Atomic pin on current block-pointer set. <1 ms, no IO.

02

Commit

WORM seal applied by the SPA layer. Immutability enforced kernel-side.

03

Replicate

Smart Replicator pushes delta-only encrypted stream to peer / vault.

04

Retire

On SLA-Policy expiry, block pointers are released back to the free list.

Creation time Minutes-hours <1 second
Storage cost Equal to source Delta-only
Tamper-evident No Yes (WORM)
Cryptographic deletion Possible Blocked by RPH
SynetoOS > Recovery Points
RecentHeldOff-siteExpired
14:23 · vm-erp-01WORM · 1m ago
14:22 · vm-crm-02WORM · 2m ago
14:21 · vm-files-03WORM · 3m ago
14:00 · daily-setHOLD · 24h
yesterday · monthlyOFF-SITE · pulled
Live Recovery Point timeline — every commit is signed, every state is queryable.
syneto.euPage 7 of 12
Capacity08
Chapter 06 · Capacity

Dedup, compression, and the math behind capacity.

SyFS applies inline deduplication and compression on the write path — the data on disk is the deduplicated, compressed form. No post-process pass, no read penalty.

SynetoOS > Pool > Capacity
180 TB
Raw
120 TB
Usable
420 TB
Logical
3.5×
Effective ratio
Reduction breakdown
Inline deduplication×1.8
LZ4 compression×1.6
Thin provisioning×1.2
Capacity tiers
Hot · NVMe (24×)38.5 / 60 TB
Warm · SATA SSD (8×)52.1 / 80 TB
L2ARC · NVMe spill1.9 / 4 TB
Effective capacity = raw × (1 − parity) × dedup × compression × thin.
Typical effective ratios

VDI / virtual desktops: 4-6×. Database VMs: 2.5-3×. Mixed VM workloads: 3-4×. Backup datasets: 5-8×. The pool report shows real-time ratios per dataset.

syneto.euPage 8 of 12
Operations09
Chapter 07 · Operations

Day-2 operations admins actually run.

Scheduled scrub — proactive integrity

A scrub reads every block in the pool, verifies checksums, and rewrites any mismatches from redundant copies. Runs weekly by default at lowest IO priority — production stays online and untouched.

SynetoOS shell
# Inspect pool health
syctl pool status hyper-prod
state: ONLINE
scan: scrub repaired 0B in 3h12m, 1 errors corrected
config: 24× nvme RAID-Z2 (no faults)
# Trigger an immediate scrub
syctl pool scrub hyper-prod --priority low
# Query Recovery Point retention for a VM
syctl recovery-point list vm-erp-01 --window 24h

Holds, retention and the SLA Policy contract

Without SLA Policy
  • Admin defines snapshots by hand
  • Retention is policy-by-convention
  • No automatic replication
  • Audit trail relies on Bash history
  • Compliance evidence rebuilt per-audit
With SLA Policy
  • Declarative tier per VM (Mission/Business/Standard/Archive)
  • Retention windows compound (60min/24h/30d/12m)
  • Replication target encoded in the policy
  • Every action emits a signed audit event
  • NIS2 / DORA reports generated on demand
syneto.euPage 9 of 12
Live dashboard10
Chapter 08 · Live dashboard

What an administrator actually sees.

SynetoOS · Pool Health · hyper-prod
OverviewDevicesRecovery PointsReplicationAlerts
128k
IOPS
0.42ms
Latency p99
8.2 GB/s
Throughput
3.4×
Dedup ratio
Vdev health
nvme0 · mirrorOK · 96% util
nvme1 · mirrorOK · 94% util
nvme2 · RAID-Z2OK · 82% util
spare-1STANDBY
Recent events
14:23:01OKRecovery Point vm-erp-01 committed (WORM)
14:22:14OKSmart Replicator delta sent: 142 MB → echo-vault
14:18:55INFOL2ARC warmup at 64% (4.2 GB/s)
13:45:02WARNDrive nvme2-bay3 temp 56°C (within spec)
12:30:00OKScheduled scrub completed: 0 errors, 0 repaired
Single-pane operations: pool health, replication, Recovery Points, alerts — same UI on every Syneto appliance.
syneto.euPage 10 of 12
Support & contact11
Chapter 09 · Support & contact

About Syneto & next steps.

Syneto designs and manufactures the complete Hybrid Cloud Ecosystem — hypervisor, data management, data protection and disaster recovery in a single plug & play platform. Founded in 2008, headquartered in Brescia and Timișoara, serving 5,000+ European businesses.

5,000+
Companies served
250+
Channel partners
18+
Years building HCI
$0
Ransom ever paid
European engineering

Designed, built and supported in Europe. ISO 9001 and ISO/IEC 27001 certified operations.

Plug & play deployment

Production-ready in under 30 minutes. No multi-vendor integration, no professional services required.

Single-vendor support

One contract, one phone number, one SLA — for the entire stack. Local language support.

Zero ransom paid

Across 18+ years of customer deployments, no Syneto customer has ever paid ransom following a cyber incident.

Single-vendor support — one contract, one phone number, one SLA for the entire stack. Three response tiers, all delivered by Syneto engineers in your language.

PlanResponseCoverageWho it's for
EssentialNBD — 8×5Software + hardwareBranch offices, test & dev
Business4 h — 24×7Software + hardwareProduction workloads
Mission-Critical1 h — 24×7Software + hardware + TAMBusiness-critical sites, DR

Talk to Syneto

Customer Support Portal

support.syneto.eu

Help Center

help.syneto.eu

Syneto Academy

academy.syneto.eu

Partner Portal

partners.syneto.eu

syneto.euPage 11 of 12

Ready to simplify?.

Join 5,000+ European companies that trust Syneto.

Italy HQ
Via Cefalonia 70
Brescia 25124
Romania
Bastion Office
Timișoara 300054
Spain
Calle Antonio Arias 6
Madrid 28009
Web
syneto.eu
syneto.eu [email protected] +39 051 095 3000
SYNETO SPA · VAT RO15999720 · REG J35/2971/2003 · ISO 9001 · ISO/IEC 27001