Computer Associates ARCserve Backup 12.5 review

The battle for the best Windows backup software has traditionally been fought between Computer Associates (CA) and Symantec with each trying to outdo the other for features. Launched at the beginning of the year, Backup Exec 12.5 impressed with its range of new features but CA's ARCserve Backup r12.5 takes centre stage as it's the first backup software product to deliver data deduplication as a standard component of the core product.

Including deduplication at no extra cost brings this technology easily within the reach of SMBs but CA goes further as it augments it with a raft of equally valuable new features. It now integrates with both Windows Server 2008 Hyper-V and VMware's VCB (VMware consolidated backup) allowing full and incremental image and file level backups of VMs to be taken. ARCserve's granular recovery also allows individual files and folders to be restored from images.

ARCserve can securely manage passwords used for encrypted backups whilst Media Assure jobs check deduplicated data at random intervals to ensure that it is not only readable but recoverable. There's role based administration and the new Dashboard provides full storage resource management (SRM) reporting facilities.

CA opted to deduplicate at the backup server as it's faster than client side deduplication and it can apply this simultaneously to multiple backup streams. It uses hash-based chunking where it segments the incoming data, examines each chunk and computes a hash value for each block of data which it stores in an index file.

If the hash value for an incoming block matches one in the index instead of storing it, ARCserve updates its reference file with the address of that block already in the data file. Post processing is also included and applied globally to all system drives, system states and virtual machines as there is a strong likelihood of duplicated data in these.

Only NTFS volumes are supported for the data and index file locations and the index should preferably reside on an SSD for the best performance. Encryption and compression can't be applied to deduplicated data but these devices can be part of a D2D2T backup strategy.

Using a Boston Supermicro dual 2.5GHz L5420 Xeon rack server running Windows Server 2003 R2, we found installation a swift affair helped along with an express option for a standalone backup server. The intuitive management console offers a backup wizard on first contact and a navigation bar pops up when requested to provide easy access to the various functions.

For deduplication you create a new backup device type and decide where the data, reference and index files are to be stored. These can be in separate locations and on remote systems as well. CA hasn't published any claims about deduplication ratios so we created our own set of tests that looked at performance for file server operations.

Using a 4GB data set consisting of 1,000 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 per cent of the data was modified in 40 per cent of the files prior to each subsequent backup. After a two week simulation we saw ARCServe deliver a deduplication ratio of 5.3:1 showing good savings on storage are quite feasible.

We also tested deduplication by backing up the system drives and system states on four Windows client systems. The total amount of data sent to the appliance was 90.1GB but our backup file store and index only had a total of 68.4GB of data resulting in an initial deduplication ratio of 1.4:1 and a storage saving of 25 per cent.