PASS - The next big thing from Hyperbac

You might have read my previous post about coming to the Hyperbac booth at PASS to see the next big thing.

So what was it that Hyperbac was announcing.

Imagine this secnario. You have your production database, if you've folowed best practice the data files will be sized for say the next 6 months which means there will be a lot of empty space in the mdf/ndf files. In addition your transaction log will be sized to be big enough to reindex your largest table and so again will generally contain a large amount of empty space. So of the 100Gb space used by the data and log files you only have 50GB of used space.

If you want to restore this database onto a dev or test system you can't just restore the data. You have to find 100Gb of space.

Hyperbac Online is the new product from Hyperbac which allows you to run your database, in read/write mode, on compressed database files. This means that you are able to restore your 100Gb database into a space that is a fraction of that space. The compression of the data will be ~70-80% (similar to what you achieve with a backup). So the 50Gb of data will compress to 15Gb, and as free space compresses to nothing you will only need 15Gb of space on your dev/test system to restore and run the database.

The initial response people have to this is what about peformance. The answer is the same that you get when talking about compression within SQL Server 2008. Even though you are compressing the data, you will be reading and writing less data and so any CPU overhead that you incurr due to the compression is outwieghed by the savings you make due to doing less IO. Hyperbac has done initial tests with a TPC-H benchmark and found that performance is better using compressed files because the IO is less with compressed files.

You might be aware that you can't run SQL on a compressed NTFS volume, in which case how is Hyperbac Online any different. Well the issue with NTFS compression is that its buffered IO and is not write ordered. Hyperbac Online isn't buffered and is write ordered and so use of Hyperbac Online ensures durability of data. There is a KB article on the requirements of an IO sub system to support SQL Server Databases Hyperbac Online meets all these requirements.

Whilst this in itself is exciting whats just as exciting is that because Hyperbac is an OS level service this compression can be applied to any file, including virtual machine vhd files. There is a demo that has a 12GB VHD compressed to 2GB.

I'm sure you can imagine the savings you can ow make across your storage. Pop over to the Hyperbac stand today and see it in action. We've also got some SQL 2008 system table maps to giveaway.

Published Thursday, November 20, 2008 2:41 PM by simonsabin


No Comments