Acronis Backup & Recovery 10

Data deduplication has without doubt been the hottest topic in the network storage industry this year, and now Acronis jumps on the bandwagon by implementing it in its disk-imaging software. Offered as an option, this feature is available only with the Advanced Server version of its new Backup & Recovery 10 software.

Acronis provides both file- and block-level deduplication, and has the ability to carry this out at both the source and target. The block-level deduplication technology is fundamentally the same as that used by the majority of other solutions.

Deduplicating at source can place high demands on host resources and to get round this Acronis offers its performance-optimised source deduplication, which creates a shorter hash value using only a small portion of each block. In this scenario the index may have the block already stored, in which case the full hash is computed and checked.

For our central management system and vault repository, we called on a Fujitsu RX330 S1 rack server with dual 2.3GHz quad-core Opterons, 16GB of memory, plus 15K SAS drives and running Server 2008 Enterprise 64-bit. Our first job was to install the central licence server and load on licence keys; you'll need a key for each managed system and another for every one you want to apply deduplication to.

Managed vaults are created within the storage node where you define the drives and folders for the vault location and its associated indexing database. During creation we also enabled data compression, but you can't encrypt a deduplicating vault.

Backup policies make light work of securing data on multiple systems as you browse the network and add each one along with global definitions on what files, folders, volumes or drives you want copied.

Acronis has no published claims about deduplication ratios, so we used our standard set of tests to find out what it could do. Using a 4.5GB data set consisting of 600 files, we introduced controlled changes within a percentage of the files during a standard backup strategy consisting of daily incrementals and weekly full backups.

After the first full backup was completed, 2% of the data was modified in 40% of the files prior to each subsequent backup. After a two-week simulation, we saw Acronis deliver a reasonable deduplication ratio of 3.4:1.

We also ran a full backup of a 44GB data volume on a remote Windows server. On completion the size of the vault and index database was 29.2GB, resulting in an initial deduplication ratio of 1.5:1 and a storage saving of about 33%.

That's the good news, and now the bad. Whenever a backup is directed to a deduplicating vault, an indexing task is fired up afterwards to populate the database with the hashes of new blocks. Acronis advised us that it does this to improve backup performance and that indexing should run at speeds of around 20GB/hour. Unfortunately, we found it could be much slower.

This wasn't an issue with our data- rate change test, as each incremental copied 1.8GB of new data and the corresponding indexing task took around five minutes. However, for the 44GB backup of the remote server, the subsequent indexing job took an unbelievable 17 hours for an average speed of 2.5GB/hour.

The biggest problem with these indexing tasks is that they lock the vault while they're running. This means you can't run further backups to that archive or recover data or volumes from it when an indexing job is running. Consequently, if your daily rate of change of data is high then Acronis' deduplication definitely isn't for you.

We've always been impressed with the level of backup, restore and disaster recovery features offered by Acronis' imaging software. You still get the same high level of features with Advanced Server 10, but the deduplication option isn't as slick as that offered free with the A-Listed ARCserve Backup 12.5 and indexing performance severely limits its appeal.

Author: Dave Mitchell