Two years after Microsoft nicked DataAllegro out from under Ingres, the open-source database vendor has delivered its hardware-loving storage and database response.
Ingres has released VectorWise, a database technology that utilizes the cache on multi-core chips to speed data crunching and analysis in massive data sets.
The VectorWise Project started in 2009 as a database spin-out from mathematics and computer scientific research hothouse Centrum Wiskunde and Informatica (CWI) after Microsoft bought DataAllegro - then working with Ingres. DataAllegro had already devised a test version of a storage engine with Ingres.
DataAllegro is now part of Microsoft's SQL Server 2008 R2 Parallel Data Warehouse built on hardware from Bull, Dell, EMC, HP and IBM. Microsoft's appliance is designed so that each node has dedicated CPU, memory, and storage, along with its own instance of SQL Server, in a parallel shared nothing architecture.
Microsoft is pitching the device as something that, by combining hardware and software, simplifies deployment and management of BI and analytics warehouses. It's the same shtick Oracle's been pushing with Exadata, with the Oracle database and Intel hardware.
Ingres is pushing VectorWise for crunching data on regular servers, without needing to buy specialized hardware optimized to work with a database and tools.
Chief executive Roger Burkhardt said VectorWise unlocks the power of what he called "modern" microprocessors without users having to configure the system themselves to optimize chip cache. He claimed performance gains of between 10 and 70 times over existing database and analytical servers running large, distributed data sets.
Burkhardt said not only does using regular servers mean lower price for customers, it means you don't have to depend on a single company for both hardware and software - something he claimed is alarming ISV customers he's spoken to when it comes to Oracle and Exadata.
"VectorWise is for customers who have a performance problem using Oracle today, have a new requirement, and know enough to know it's going to be extremely complex to solve it with Oracle," Burkhardt said.
Ingres is also pitching VectorWise as working with NoSQL, the data movement building large-scale data processing breaking with traditional SQL norms of storage for size and performance.
VectorWise works with Hadoop via the HadoopDB project, which will parse an SQL statement and make it run over 50 computers but act like it's on a single machine. The idea is you get the scale of Hadoop while retaining the familiarity and performance of working with SQL.
Daniel Abadi, director of Database Research at Yale University, where HadoopDB was invented, said in a statement that VectorWise underpins large-scale HadoopDB applications.
VectorWise is available for download, and priced on a subscription basis according to your number of cores. Two sockets with two cores starts at $24,000 per year, with discounts provided for volume, Burkhardt said. ®