At the Teradata Universe in Barcelona, Scott Gnau, Chief Development Officer unveiled Teradata’s latest flagship product the Active EDW 6680. With support for SSD and SAS drives along with just under 1TB of RAM, the Active EDW 6680 is described by Gnau as being Teradata’s answer to the need for speed in high end data warehouses.
The 6680 comes with its own intelligent software is fully automated with no direct DBA intervention allowed. The goal is to remove the DBA from the need to be constantly optimising indexes and tables for performance and this is achieved by the built in self managing and self optimising management software. According to Gnau, the 6680 will provide customers with a 4x performance per floor tile inside their datacentre.
What makes the 6680 different from other solutions is the way that it optimises data into three different zones, hot, warm and cold. According to Gnau, when you analyse the way that data is used in data warehouses, you discover that 43% of all IO is focussed on just 1.5% of the data stored. As you move forward, 85% of the IO address just 15% of data and 94% of IO touches no more than 30% of data in the data warehouse.
The result is that significant amounts of data rarely, if ever, used. In order to optimise for the data that is used, the 6680 will store the most used 20% of data on its SSD drives. This approach is very different to that of Teradata’s competitors who currently utilise SSD as an extension of RAM to create large in-memory datasets. The problem of this, from Gnau’s perspective, is that it doesn’t really optimise the use of data at all and just keeps a lot of unused data on high value SSD.
The 6680 splits the remaining 80% of data into two groups of roughly 40% each called warm and cold. Warm data will be that which has recently been in the hot zone and cold data will be data that has not been accessed for some period of time. At present, Gnau says that there are no plans to start doing predictive analysis of the data to see what data might become hot and move it to the warm zone. While not ruling this it out completely, he said that much would depend upon customer demand.
Gnau also claims that the movement of data in and out of the hot zone will not impact the performance of the 6680. This is because the decision making and movement will be done using spare IO and processing cycles so the users will not see any impact on performance at all.