Full Post
Please also read:
Table Comparison: Transparent RAID vs. RAID over File System
Most users should choose Transparent RAID (tRAID) over RAID over File System (RAID-F). This is because tRAID covers a much wider range of deployment scenarios than RAID-F. For the sake of discussion, we will dismiss the real-time feature of RAID-F since it is still experimental, and we will only focus on its Snapshot RAID feature. Snapshot RAID requires users to split their data and separate data that changes often from data that changes rarely. Over the years, we have found that users prefer a single aggregation system. That is, most want a system that will combine all of their data regardless of type, protect such data, and finally pool the data into a single volume for a unified data access. Transparent RAID better meets those criteria than RAID over File System.

That’s not to say RAID-F does not have its place. For users that can separate their static/semi-static data from data that changes frequently, a lot of efficiency can be gained from running a Snapshot array. If the bulk of your data is comprise of disc images, movies, music, pictures, zip archives and the likes, backup images, etc., it might make sense to create a Snapshot array that only hosts such data. Then, you could put your dynamic data on a RAID 1 array if it can fit on a single disk (if your data can fit on a single disk, the best protection is to simply mirror it) or on a tRAID array if it spans more than one disk.

Snapshot RAID is more efficient than any other RAID form simply because parity updates happen at scheduled times rather than in constant real-time. When you write to a disk that is part of a Snapshot array, only that disk is active. In contrast, in tRAID, both the disk being written to and the parity disk(s) will be active. Further, tRAID takes a hit in write speeds due to the overhead in real-time parity updates.

One would choose RAID-F over tRAID:

  • When not protecting whole disks as only RAID-F can protect specific folders or files or network shares. Transparent RAID can only protect whole disks. RAID-F can protect data that resides over the network and can even excludes data on a disk from protection.
  • When wanting to protect USB disks. Although USB disks can be made part of a transparent array, doing so is highly discouraged.
  • When wanting to create a Storage Pool volume that is not strictly tied to the parity array. In RAID-F, one can create multiple Snapshot arrays and combine them under a single Storage Pool. Note that a RAID-F pool can include tRAID disks.
  • When needing to move disks freely from system to system without impacting the parity data. Since RAID-F only protects files rather than the disks the files are on, parity is compromised only when such protected files are edited. In contrast, if you access a tRAID disk outside of a tRAID array, the parity will be compromised as your OS will silently write to the disk.
  • When the data to be protected is exclusively static or changes very rarely. In this particular case, Snapshot RAID is simply more efficient.
  • When a better write throughput is desired. Transparent RAID takes in hit in write speed due to the real-time parity update overhead.

One would choose tRAID over RAID-F:

  • When the data to be protected is a mix of data that changes frequently and static data. Snapshot RAID is not appropriate for data that changes frequently. Furthermore, the real-time feature of RAID-F is experimental and does not deal well with fragmented data. Transparent RAID has no such issue.
  • When live data reconstruction is a must. When a drive fails in RAID-F, that data is unavailable until it is restored. In contrast, a disk failure is fully transparent in tRAID as the failed data is reconstructed live as if no drive had failed.
  • When minimal maintenance is required. Snapshot RAID requires frequent parity sync to pick up data changes. Although this is automated through the scheduler, there is a greater consciousness in having to keep the parity data up to date and valid.
  • When simplicity is desired. Transparent RAID is a much simpler platform.
  • When centralized management of multiple systems is desired. Transparent RAID features a centralized management UI that can manage an unlimited number of host systems. This includes centralized SMART monitoring, task execution, and notification.

Ultimately, you should trial both RAID-F and tRAID to truly decide which one fits you best.