Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

Reverse incremental vs forward incremental backup: which restore point is fastest and what each costs at night

Technical article
Backup & Recovery

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

Forward incremental keeps a full backup and adds each night's changes after it, so the newest point is the full plus the changes that followed. Reverse incremental does the opposite: each night it folds the changes into the full, so the newest point is always a complete full, and it keeps small "undo" files that step back to older points. Reverse restores the newest point fastest and older points slowest, and it puts the heaviest work on the backup store each night.

If you are choosing today, check one fact first. In Veeam Backup & Replication version 13 the reverse incremental option "is deprecated, and you cannot select it for new jobs", although existing jobs "will continue to work after upgrading" (Backup Methods).

What does each method store?

Veeam describes the forward chain as "the first full backup file (VBK) and a set of forward incremental backup files (VIB) following it". For reverse it is "the last full backup file (VBK) and a set of reverse incremental backup files (VRB) preceding it" (Reverse Incremental Backup).

In a reverse chain each night's changed blocks are injected into the full to rebuild it to the latest state. A reverse incremental file is created "containing data blocks that are replaced when the full backup file is rebuilt". "As a result, the most recent restore point in the backup chain is always a full backup, and it gets updated after every successful backup job session."

Forward incremental Reverse incremental
Full file Older, fixed until the next full Always the latest state, rebuilt nightly
Increment files Changes after the full Undo files that step back from the full
Newest point Full plus the changes after it The full itself
An older point Full plus changes up to that point The full, with undo files applied back to that point

Which restore point is fastest?

Veeam's comparison is direct. For reverse: "Restore to the latest restore point ... is the fastest compared to other methods", and "Restore to earlier restore points is slower than for other backup methods". To reach a particular older point it "applies the required VRB files to the VBK file".

For forward chains with periodic fulls, Veeam says restore is "the most optimal in time compared to other methods (in cases when you do not restore to the earliest or latest restore point)", because the chain is divided into short series of full backup files and incremental files. Forever forward chains make the earliest point the fastest, because the first available point is always a full.

Restore target Reverse Forward with periodic fulls
Newest point Fastest Depends on how far it is from the latest full
A point in the middle Slower, more undo files to apply Documented as the most optimal for such points
The oldest point kept Slowest Starts from a full

What does each cost at night?

Veeam ranks the load on the backup storage. Forward incremental "produces the lowest I/O impact on the backup storage", though it rises on synthetic full days. Reverse incremental "produces the heaviest I/O impact ... compared to other backup methods", because the job "injects changed data blocks into the full backup file and also creates reverse incremental backup files".

Example (illustrative): a 1 TB server app01 changes 5% a night, so 50 GB of blocks change. Decimal units: 1 TB = 1,000 GB.

  • Forward: the repository writes one 50 GB increment file.
  • Reverse: the repository writes the 50 GB of new blocks into the full, and also captures the 50 GB of old blocks they replace into an undo file. Moving the old blocks means reading them from the full and writing them out. Veeam's Best Practice Guide gives the ratio: "1 x write I/O" for forward incremental and "3 x I/O (1x read, 2x write)" for reversed incremental (Backup job methods). Here that is about three 50 GB operations (150 GB) against one (50 GB) for forward.
  • At an assumed sustained 150 MB/s of repository throughput, 50,000 MB / 150 MB/s is about 333 s (5.6 minutes) for forward, and 150,000 MB / 150 MB/s is 1,000 s (16.7 minutes) for reverse.

The throughput figure is an assumption, and it is applied equally to every operation, which is itself a simplification. The same Veeam table lists forever forward incremental at "3 x I/O (1x read, 2x write)" too, so reverse is not the only method with that cost. It also marks reversed incremental as benefiting from Fast Clone (ReFS or XFS block cloning), which can reduce the extra I/O on a repository that supports it.

Veeam also notes that these transformations "can lead to the fragmentation of the full backup file", so both reverse and forever forward chains are given a scheduled compact. The compact "creates a new full backup file in the target repository" and needs space "to store a file of the full backup size" (Compact of Full Backup File).

What does retention do to each?

Reverse is tidy. "Veeam Backup & Replication immediately deletes the earliest reverse incremental backup file as soon as it becomes outdated" (Reverse Incremental Backup Retention Policy). The space stays close to one full plus the undo files for the kept period.

Forward is not, and Veeam explains why. To restore you need "a full backup file and a chain of subsequent incremental backup files on disk", and the product "does not create full backups automatically based on retention limits". So "on some days, there will be more restore points on disk than specified by retention policy settings". In its three-day example the first full "is not deleted" on the day the retention is reached, and the whole old chain goes only after the second chain has aged enough (Forward Incremental Backup Retention Policy).

Forward therefore needs more space than a simple count suggests. Veeam says forward incremental "requires more storage space than other methods because the backup chain contains multiple full backup files".

What happens if something is damaged?

The two shapes fail differently.

  • Forward: "If you delete a full backup file, the entire chain of incremental backup files will become useless". Veeam's reason for periodic fulls is that splitting the chain "lowers the chances of losing the backup chain completely and makes this backup method the most reliable" (Forward Incremental Backup).
  • Reverse: one file is rewritten every night and every older point is derived from it. By the same logic as the forward case, damage to that file affects every point that depends on it. Veeam does not state this; it follows from the design.

Veeam's health check for reverse chains "always verifies only the latest restore point in the backup chain, which is always a full backup file". It does not check the older points by itself (Health Check for Backup Files).

How can I check this on my own system or with my provider?

  1. Ask which method each job uses, and whether the label is Veeam's. Other products use "reverse incremental" differently, so ask what is stored and what a restore of the oldest point reads.
  2. Time two restores, not one. Restore the newest point and the oldest point to a scratch location and record both times. A provider quoting only the fast one has told you half.
  3. Ask when the last compact ran and what it needs. It needs free space of about one full.
  4. On a Veeam server running version 13, open the job's settings. The reverse option cannot be chosen for a new job, so seeing it on an existing job means that job was created earlier.
  5. Ask how many points are on disk on a day just before a full. For a forward chain this is the number that shows real space use.

Risks and limits

  • Risk: switching an existing job between forward and reverse changes the chain and the storage load. Veeam says a switch "does not transform the previously created chain" and instead "creates a new backup chain next to the existing one"; switching from forward to reverse first creates a new full backup file, which needs space (Switching Between Backup Methods). Backing out means switching again, which creates another chain; for a switch between forever forward and forward, the page says the old chain is deleted when the new chain reaches the retention limit.
  • Risk: deleting backup files by hand, in either method, can break every restore point that depends on them. There is no way back unless another copy exists.
  • Veeam is the only product described here. Timings and I/O comparisons are Veeam's own relative statements, not benchmarks.
  • Because reverse is deprecated from version 13, plan a move to another method before relying on it for a new design.
  • The 5% change rate and the 150 MB/s throughput are assumptions for the example.

NetLeverage includes a daily full disaster-recovery backup of each hosted server at no extra charge. Talk to an engineer.

Sources