This article is more than 1 year old

Flash biz and virtual database chum: Our tests show we bust spinning rust

Pure Storage and Delphix splatter platters in home-made benchmark

A flash array speed merchant and database virtualisation firm have got together to run a benchmark that shows their disk array brethren eating their dust.

Delphix takes a master database and makes virtual copies of it for whomever needs them, including accounting types, big organisations trying to control data spread, and testers and developers. The virtual copies consume far, far less space than full clone copies yet are just as functional. What's not to like?

Pure Storage, a flash array startup which has just bagged a cool $150m to build out its business and make better kit, has a core message about apps accessing stored data at NAND flash speed instead of using lots of oh-so-slow disk spindles – and at an affordable cost as well, since the boxes do inline deduplication.

The two have got together and compared database copies running on "a disk-based storage array from a leading legacy storage vendor" to a Delphix/Pure Storage double act using virtual copies on a flash array.

The disk array ran 26 copies of a database, each a terabyte in size and supporting 35,000 transactions a minute (TPM), 910,000 TPM overall, with an array cost of around $2m.

Our disk array dissing duo then ran the same 26 databases as a master with virtual Delphix copies on a Pure Storage flash array. Instead of a 26TB-plus capacity array they only needed 1.5TB of flash capacity in total, and supported around a million TPM at a cost of $200,000.

For 10 per cent of the cost and six per cent of the capacity of the disk drive array, they delivered more database TPM, and that's without saying anything about lower requirements for physical space, electrical power and cooling savings.

The two state: "Virtual flash databases can be created, refreshed, or rolled back in minutes, via self-service, by development or test teams. Block sharing across copies enables creation of large databases without duplication or movement of data across networks or arrays."

It may be a home-brewed benchmark, but it's food for thought. ®

More about

TIP US OFF

Send us news


Other stories you might like