The Netherlands-based company backs up office productivity and SQL database information as well as Adobe Photoshop and Illustrator-based artwork for clothing and footwear designs. But backups – using Commvault direct to tape – were encroaching on the business day and some data was being excluded from jobs, so that the most business-critical data could be squeezed into the available window.
Because Waalwear was already a user of Data Domain products, it made plans to roll out their use across the sportwear firm's European sites. The O'Neill IT team oversaw the implementation of one Data Domain DD565 appliance with an expansion unit at the firm's Warmond headquarters, plus four DD510s at four other sites in Holland, Belgium, German and France. A sixth unit – a DD565 – will be added at a disaster recovery site in the next three months.
Maljaars declined to reveal the cost of the implementation but a DD565 costs £72,652, the DD510s £13,680 each and the expansion unit £27,360 at current UK prices
The key benefits for O'Neill are a massive reduction in backup times as well as being able to keep copies on-site for up to a year, eliminating the need for time-consuming physical retrieval of tape and its restoration. O'Neill is backing up 57TB of data to just 3TB of disk space using a compression ratio of 18:1
"Backup is now taking only two hours, as we can do incrementals through the day also, says Maljaars. "We can also restore in four or five hours straight from Data Domain instead of having to phone up for tapes and get them delivered, which took a whole day before."
Paradoxically, O'Neill is also backing up more data than before it implemented the Data Domain appliances, because now it can copy all data, instead of being forced to choose what it can squeeze into the backup window. "We're now backing up more data than before – 5.3TB instead of 1.4TB," says Maljaars. "That's because previously we just couldn't back everything up, as it would have taken 20 or 30 hours if we included, for example, all the image files which are important but we can do without. "So we used to have to decide what was most important. Now we can back everything up and not have to make that decision."
More and more companies are turning to data deduplication for their expanding backup needs. It is essentially a form of compression, with duplicate blocks of data in a set identified and replaced with some form of tag pointing to the full copy. Deduplication comes in two forms: inline, in which data is deduplicated as it is copied, and post-process, in which deduplication takes place after copying to the appliance.
The technology's suitability for particular use cases will depend on the data profile of an individual user. Those with a high volume of large and duplicate files will benefit the most. Those with a high proportion of small and unique files will gain least. At the same time, users must consider whether their deduplication needs will need to scale and whether diminishing returns are likely on this. Backup is the natural use case for data deduplication, says Simon Robinson, research director with analysts The 451 Group.
"Backup is where you get the most repetition in data traffic, with minimal deltas between backup sets," Robinson says. "That's where the majority of opportunity lies for these products. It is not viable for primary storage because of the latency that results from the necessary overhead which goes with the deduplication process."
Robinson also points to some risks in using data deduplication tools that result from the way it disassembles data. "You should be very certain there will be no data integrity issues before going down the dedupe road because of what the process does to data," he says. "Organisations that need to prove data is un-tampered with or that need to encrypt backed-up data will need to be certain the product they use will be compatible with these requirements."
When Storage magazine conducted a survey on data deduplication earlier this year, it found that 79% of those questioned were not yet using data deduplication and of those 46% aimed to wait for the technology to mature. A key concern raised in the survey was whether deduplication would add extra layers of complexity to the restore process. More than half of respondents (53%) had plans to implement data deduplication over the next two years, while 37% had no plans.