Taking the Analytics Pressure Off the Data Warehouse

Share it on Twitter  
Share it on Facebook  
Share it on Linked in  

It turns out administrators of data warehouses are spending 80 percent of their time trying to manage a particular class of application workload. According to John Santaferraro, vice president of solutions and product marketing for ParAccel, a provider of a columnar database that runs on a massively parallel processing (MPP) engine, data warehouses are staring to choke on ad hoc queries being generated by analytics applications. That in turn is putting a lot of pressure on data warehouse administrators to speed the performance of those applications.

To address that specific issue, ParAccel announced today that it is has created ParAccel Analytic Offload Solutions for Oracle and Teradata that use ODBC interfaces to identify and then move analytic workloads from data warehouses to that ParAccel Analytic Platform that Santaferraro says is substantially less costly to own and manage than either an Oracle or Teredata database, while still being about 100 times faster.

The main business issue, says Santaferraro, is that at a time when there is a critical shortage of analysts skilled enough to build and run analytics applications, organizations should not be tying those people up by making them wait hours, sometimes days, to run a query against a data warehouse based on a relational database. In contrast, a columnar database is not only optimized for analytics applications, it can execute those queries on an MPP platform in a matter of seconds. That means that the number of analytics tasks that can be performed by any given analyst exponentially increases, says Santaferraro.

Santaferraro says that instead of thinking in terms of physical data warehouses, IT organizations should be moving to logical data warehouse models that span different technologies that are optimized for different types of workloads, algorithms and data types. That new data ecosystem, he says, should encompass relational, columnar, NoSQL databases alongside open source data management frameworks for unstructured data such as Hadoop.

Of course, finding people skilled enough to manage all that is still a challenge. But as analytics applications continue to become more strategic, it’s inevitable that a one-size-fits-all strategy based on relational database technologies simply isn’t going to be able to keep pace with data management requirements that are now becoming more challenging with each passing day.