Skip to main content
infrastructure

Storage Architecture

Storage topology, capacity planning, and data distribution across all NAS units and hypervisors

February 23, 2026

Storage Architecture

Total raw storage across the infrastructure exceeds 252TB, spread across four NAS units on two physical sites. The storage hierarchy is tiered: Meridian-Host handles active workloads and media serving, Cassiel-Silo provides high-capacity secondary storage and backup, Mobius-Silo holds archived cold data, and Rigel-Silo on the Milky Way is in an uncertain state (potentially dead hardware).

Storage Inventory

HostIPSitePlatformCapacityRoleStatus
Meridian-Host192.168.20.50AndromedaUnraid~100 TBPrimary NAS, Docker hostActive
Cassiel-Silo192.168.20.7AndromedaSynology DSM120 TBSecondary / backupActive
Mobius-Silo192.168.20.170AndromedaSynology DSM22 TBArchive / cold storageActive
Rigel-Silo10.42.0.10Milky WaySynology DSM~10 TBLocal NASUnknown / potentially dead

Total: 252+ TB across all systems (242 TB confirmed active on the Andromeda, ~10 TB uncertain on Milky Way).

Meridian-Host (Primary NAS)

IP: 192.168.20.50 Tailscale: 100.64.0.15.30 Platform: Unraid Capacity: ~100 TB CPU: 24 cores Site: Andromeda

Meridian-Host is the central storage and compute node. It runs Unraid, which provides a parity-protected array with the flexibility to mix different drive sizes. The ~100TB figure represents usable array capacity after parity.

Storage Layout

Unraid’s storage model differs from traditional RAID:

  • Array: Multiple data drives protected by 1-2 parity drives. Drives can be different sizes. Parity is calculated across the array but each drive contains a complete filesystem.
  • Cache: One or more SSDs/NVMe drives used for write caching and Docker/VM storage. The cache pool is typically a BTRFS mirror.

Primary Uses

  • Media libraries: Movies, TV shows, music, audiobooks. This is the source for both Plex instances running on Polaris-Media (192.168.20.201).
  • Docker volumes: Persistent data for all Docker containers running on Meridian-Host (Tautulli, Audiobookshelf, Grafana, etc.).
  • Build artifacts: Binary packages from the dr-Meridian-Host build drone (24 cores).
  • Downloads: Active downloads via ruTorrent before sorting into media libraries.
  • Shared storage: NFS/SMB exports consumed by other hosts on the Andromeda, including the Polaris-Media LXC on Tarn-Host.

Shares

Meridian-Host exports shares via SMB and NFS. Key shares include:

ShareProtocolConsumersContent
mediaNFS/SMBPolaris-Media, ShieldMovies, TV, music
audiobooksNFS/SMBAudiobookshelf containerAudiobook library
downloadsSMBLocal clientsActive downloads
appdataLocalDocker containersContainer persistent data
backupsNFS/SMBOther NAS unitsBackup targets

Unraid Web UI

http://192.168.20.50

The Unraid web UI provides array management (start/stop, disk assignments, parity checks), Docker container management, share configuration, and plugin management. It is the primary interface for managing Meridian-Host’s storage.

Cassiel-Silo (Secondary NAS)

IP: 192.168.20.7 Platform: Synology DSM Capacity: 120 TB Site: Andromeda

Cassiel-Silo is the largest single storage node in the infrastructure at 120TB. It runs Synology DSM and serves as secondary storage and a backup target for critical data from Meridian-Host.

Primary Uses

  • Backup: Receives scheduled backups from Meridian-Host for critical data (media libraries, Docker volumes, configuration files).
  • Secondary media storage: Overflow media storage when Meridian-Host’s array is full or for content that is less frequently accessed.
  • Large archive: Bulk data that does not need the performance of Unraid’s cache tier.

Synology Features in Use

  • Hyper Backup: Synology’s backup tool for scheduled backup jobs.
  • Shared Folders: SMB/NFS exports for network access.
  • DSM Web UI: http://192.168.20.7:5000 for management.

Backup Relationship with Meridian-Host

Meridian-Host (100 TB, Unraid)

    ├── Critical data ──────► Cassiel-Silo (120 TB, Synology)
    │   (media, configs)       Hyper Backup / rsync

    └── Archive data ──────► Mobius-Silo (22 TB, Synology)
        (cold storage)         Manual / scheduled copy

Cassiel-Silo has more raw capacity than Meridian-Host, making it suitable as a backup target with room to grow. The 120TB pool can absorb the full contents of Meridian-Host’s critical shares with headroom.

Mobius-Silo (Archive NAS)

IP: 192.168.20.170 Platform: Synology DSM Capacity: 22 TB Site: Andromeda

Mobius-Silo is the cold storage tier. Data that needs to be retained but is rarely accessed goes here. Think old project files, completed media collections, historical backups, and anything that does not need to be on a fast array.

Primary Uses

  • Cold archive: Data retained for long-term storage.
  • Historical backups: Old backup snapshots rotated off Cassiel-Silo.
  • Completed collections: Media or data sets that are “done” and unlikely to change.

Access

http://192.168.20.170:5000   # Synology DSM Web UI

Mobius-Silo is accessed infrequently. It is on the Andromeda LAN and reachable via Tailscale from any peer, but day-to-day operations rarely touch it.

Rigel-Silo (Local NAS)

IP: 10.42.0.10 Platform: Synology DSM Capacity: ~10 TB (estimated) Site: Milky Way

Rigel-Silo is the only storage unit on the Milky Way. It was used for local media storage and ruTorrent hosting before those workloads migrated to the Andromeda site. Its current status is uncertain — it may have hardware failure and needs physical inspection.

Status: Unknown / Potentially Dead

The unit has not been responsive recently. Possible causes:

  • Drive failure (aging hardware)
  • Power supply issue
  • Network configuration change
  • Simply powered off

Recovery plan: Physical inspection, check power and network, attempt DSM access at http://10.42.0.10:5000. If hardware is dead, salvage drives for data recovery if needed.

Historical Role

Before the Andromeda infrastructure matured, Rigel-Silo handled:

  • Local media storage for the Milky Way
  • ruTorrent download management
  • SMB shares for the workstation

These functions have since moved to Meridian-Host and the Andromeda.

Data Distribution by Category

Media Libraries

TypePrimary LocationBackup LocationSize (est.)
MoviesMeridian-HostCassiel-Silo~40 TB
TV ShowsMeridian-HostCassiel-Silo~30 TB
MusicMeridian-HostCassiel-Silo~2 TB
AudiobooksMeridian-HostCassiel-Silo~1 TB
Total mediaMeridian-HostCassiel-Silo~73 TB

Infrastructure Data

TypeLocationSize (est.)
Docker volumes (Altair-Link)Altair-Link local~50 GB
Docker volumes (Meridian-Host)Meridian-Host cache~200 GB
Build artifacts (swarm)Orchestrator hosts~20 GB
Proxmox VM disksIzar-Host, Tarn-Host local~500 GB
Git repositories (Gitea)Altair-Link~10 GB

Backup and Archive

TypeLocationSize (est.)
Media backupsCassiel-Silo~73 TB
Config/system backupsCassiel-Silo~500 GB
Cold archiveMobius-Silo~15 TB

Capacity Planning

Current Utilization

Meridian-Host:      ~100 TB usable  │  ~85% utilized  │  ~15 TB free
Cassiel-Silo:      120 TB usable  │  ~60% utilized  │  ~48 TB free
Mobius-Silo:   22 TB usable  │  ~68% utilized  │  ~7 TB free
───────────────────────────────────────────────────────────────
Total:           242 TB usable  │  ~70% utilized  │  ~70 TB free

Growth Projections

Media acquisition is the primary driver of storage growth. At current rates:

  • ~2-4 TB/month new media content
  • Meridian-Host has approximately 4-8 months before requiring new drives or offloading to Cassiel-Silo
  • Cassiel-Silo has significant headroom (~48 TB free)

Expansion Options

  1. Add drives to Meridian-Host: Unraid makes it trivial to add drives. Pop in a new drive, assign it to the array, done.
  2. Offload to Cassiel-Silo: Move less-accessed media from Meridian-Host to Cassiel-Silo to free array space.
  3. Replace smaller drives: Swap older smaller drives in any unit with larger ones.
  4. New NAS unit: If Milky Way needs local storage, replace Rigel-Silo or deploy a new unit.

Cross-Site Access

All storage is accessible across sites via Tailscale:

  • Meridian-Host NFS/SMB shares are reachable from Milky Way via 100.64.0.15.30 or through the Andromeda subnet route (192.168.20.50).
  • Cassiel-Silo and Mobius-Silo are reachable via their Andromeda LAN IPs through Tarn-Host’s subnet route.
  • Performance for cross-site file transfers depends on internet bandwidth at both sites. Large transfers (multi-TB) should be planned during off-peak hours.

Disaster Recovery Considerations

Single Points of Failure

  • Meridian-Host: If Meridian-Host’s array fails, primary media and Docker workloads are down. Cassiel-Silo has backup copies of critical media but not Docker volumes.
  • Cassiel-Silo: If Cassiel-Silo fails, backup copies are lost but primary data on Meridian-Host is unaffected.
  • Mobius-Silo: If Mobius-Silo fails, archived data is lost. Some of it may exist on Cassiel-Silo or Meridian-Host as older copies.

What Is Not Backed Up

  • Docker container configurations on Altair-Link (should be in Git but may not be fully captured)
  • Proxmox VM disk images (no automated backup)
  • Build swarm state and binary packages (transient, rebuildable)
  • Rigel-Silo data (unknown status)

Recovery Priority

  1. Meridian-Host — Highest priority. Restore Docker services and media access.
  2. Plex services — Restore Plex instances (depends on Meridian-Host media being available).
  3. Cassiel-Silo — Restore backup target availability.
  4. Mobius-Silo — Lowest priority, cold storage can wait.
storagenasunraidsynologybackup