Ordnance Survey supplies a wide variety of leisure, commercial and government customers with mapping and digital geographic information. The information systems department has a weekly back up requirement of 60 terabytes of cartographic, photographic and administrative information and has to guarantee longevity of that data in its archives.
SearchStorage UK: Are there any prominent storage technologies you will not invest in next year?
Dave Lipsey: I loathe spending money on tape and wouldn't see another penny go to it if I could help it. It is unreliable, difficult to handle and I'd like to completely get away from it if I could, but we can't. We may upgrade to LTO4 but if I can find a way of avoiding it, I will.
What proportion of your budget is accounted for by storage, and is that amount increasing or decreasing?
Lipsey: As an organization 90% of our revenues are earned from digital data we supply to our customers so storage and data longevity is very important to us. Around 12 percent of our total annual budget is accounted for by IT infrastructure – not including staff costs – and around half that is our storage budget, which is around £1 million.
The budget is certainly increasing chiefly because people are specifying better systems. Our customers want their data assured so we have to have systems that perform and are highly available and if they go down are quick to recover. If we can save money on electricity, power, cooling and space that's what we will be spending it on.
In terms of spending, what are your key storage priorities for the coming year?
Lipsey: One of our main ongoing projects is to improve recovery time by moving our 30-day back up from tape to a Data Domain disc-based system with data deduplication. As much of our data is very similar we can achieve as much as a 50:1 compression ratio. We began the implementation in July and now about a quarter of our daily production data is dealt with in this way. We will move about 75 percent of data to Data Domain eventually. The main reason for the move is because of tape being unreliable and the amount of manual handling we have had to do with it.
It has also improved recovery times. For a small file, single restores are nearly instantaneous whereas it used to take anything from five minutes to 30 minutes to restore a single file from tape. Even larger files can be restored twice as fast.
For our aerial imagery archives we're writing to Plasmon UDO II 60Gbyte write-once-read-many optical discs using Bridgehead HT Filestore software. Dual copies are made of all data with one copy going to long term archiving while the other resides on the online archive - to which internal customers have access - for one year.
The next phase will see most of the rest of our data written to disc. We can justify it in terms of cost because it means we are able to cut down on tapes and improve security – now we have the ability to send data across the network to our disaster recovery sites rather than have vans trundling down the motorway.
There will always be about 10 percent of our data that we have to keep on tape. These are TIFF format aerial photography images - which can't be deduplicated – and are used as the basis for mapping as well as imagery sold in its own right.
Any other storage plans for 2008?
Storage consolidation is another project we have planned. We have low utilisation and lots of different arrays and we'd like to virtualise across the whole system. In the past we've bought arrays as and when we've needed them and it has seemed a good idea at the time but a few years down the line we're not achieving good usage. This project is in its early stages, though, and we are working out what's possible.