Product Reviews

Data de-duplication is going to be a hot technology this year, as it offers a clear solution to businesses struggling to cope with exponential rises in demand for network storage. Traditionally, de-duplication products have been priced up for enterprises with very deep pockets, but HP's latest D2D4000 Backup System family delivers the first SMB solution.

The D2D4000 brings together a smart combination of virtual tape libraries (VTLs) and HP's own de-duplication technology. This 2U appliance provides a D2D backup and data-restoration solution and can present up to 16 VTLs over iSCSI and FC, allowing it to support as many servers for simultaneous backup operations. It offers four different library emulations while choices for virtual tape drives extend to LTO-2, LTO-3 or LTO-4. Each VTL can be configured with up to two tape drives and a total of 24 cartridge slots.

We'll start with a closer look at HP's de-duplication technology. There are a number of different methods currently available and HP opted for dynamic de-duplication, which is carried out at the appliance as it receives the backup data. It computes hashes for each 4K block, or chunk, of data that uniquely identifies it and stores them in an index on the appliance.

As data is sent to the appliance it compares their hashes with those already stored, and if they match it simply updates the reference pointer on the existing block. If the hash value of the block doesn't match anything in the index then the block is stored on the appliance and its hash value is added to the index. Dynamic de-duplication using a hash index doesn't require any client software loaded on the host systems. Furthermore, it can work with any type of data format and, after de-duplication has been performed, the appliance applies compression during its write phase.

In this exclusive review, we look at the D2D4004i dual-port iSCSI model, which comes with six 750GB SATA hard disks preconfigured in a RAID6 array. The 4000fc models are equipped with dual-port FC HBAs, but the appliance can support both connection types simultaneously. Installation is very well handled, as a separate wizard routine locates the appliance on the network, installs the LTO tape drivers, and downloads the latest Microsoft iSCSI initiator. You choose your library plus the number of slots and tape drive models,

ADVERTISEMENT

and the appliance is then configured automatically.

The tidy web interface opens with plenty of operational information and allows you to manually create VTL targets and associate them with specific hosts. As new initiators log on, the appliance can automatically create a VTL for them along with bar codes for each cartridge. Off-site storage is covered, as the Tape Attach menu allows data from selected cartridges to be copied or archived to a physically attached tape drive, and also imported into the appliance from a cartridge as well.

For testing, we wanted to see whether HP's claimed de-duplication ratios of 20:1 to 50:1 were achievable in the real world. We created a set of three automated tests that looked at performance for a file server, SQL Server and Exchange server. We created 4GB data sets and introduced controlled changes to them at the file, database row and email levels. Our testing hardware comprised a Dell PowerEdge 2900 quad-core Xeon domain controller running Windows Server 2003 and Exchange Server 2003. Symantec Backup Exec 12 (web ID:195450) managed backup operations, and we installed this on a Dell PowerEdge 1950 quad-core Xeon system running Windows Server 2003 file services and SQL Server 2005.

The tests simulated a three-month backup period, using a standard cycle of weekly full and daily incremental or differentials as dictated by best practices for each application. After the first full backup was completed, controlled changes were made to the data prior to each backup and reports gathered on the amount of data sent to the appliance and how much was actually being stored on it.

For the file server test we saw ratios of between 22:1 and nearly 70:1, although the latter was achieved when we made small changes to all of the test files. Changing 60% of the files prior to each backup returned a ratio of 38:1. For the Exchange backup test, we saw the ratios range from 30:1 to more than 40:1. The SQL Server results were also above the minimum quoted rate, as we saw these range from 21.5:1 to over 24:1.

Of equal value were the storage savings we realised at the end of the tests. With a 4GB data set used in all three application tests, we saw savings of between 70GB and 229GB for the file server backups after three months. Exchange Server saw savings of between 156GB and 176GB, while the SQL Server tests returned savings of 52GB to 59GB.

The combination of dynamic de-duplication and data compression clearly delivered the goods in our real-world tests. Along with big savings in storage, the D2D4000 Backup System offers an excellent range of VTL features, it's very easy to deploy, and is the first solution to bring this new technology within easy reach of the SMB.