About our new reference architectures. When DATAllegro joined Microsoft last fall, we saw an opportunity to help SQL Server customers leap forward in scalability and performance in a short time frame, by creating new balanced configurations specifically for data warehouses.
We initiated the SQL Server Fast Track Data Warehouse offering and within the space of just a few months, we had tested and proven several new reference configurations on HP and Dell servers, as well as EMC storage. In addition our team drafted a pair of white papers – an introduction and an implementation guide.
Several things make Fast Track unique and compelling. At first glance, these reference architectures provide orders of magnitude improvements in speed, with more predictable performance. The cost per terabyte is lower than most other data warehousing reference architectures for SQL Server, and as of this date, much lower than many DW appliances on the market. Customers that have EA, APA or other licensing agreements in place enjoy even lower pricing.
Going one level deeper, there are a few critical principles that guide us in building a new reference configuration. These are the first reference configurations developed and tested by Microsoft, and further, they are designed with a laser focus on optimizing for data warehousing, not all-purpose. Sequential IO results in far better performance than random IO, so we avoid random IO as much as possible.
We’ve strived to balance the utilization across all components, from the CPU to the SAN storage disk drive, which is also unique. For example, there’s often a tendency to max out spending on processors, but in practice the processors are underutilized when other components are running at 100%. With a Fast Track DW system, every component operates at or near its maximum or optimal rate, thus eliminating hardware performance bottlenecks.