By Isaac Moreno Navarro
With the advent of the data warehouse, the possibility of using the same infrastructure for online transaction processing (OLTP) as well as for online analytical processing (OLAP) has become a controversial subject. You will find that different database vendors have different points of view.
Let’s explore this topic by starting with a brief explanation of OLTP and OLAP.
Online transaction processing (OLTP)
This is the traditional way databases used to work – by placing small queries against a low number of rows and retrieving information. For instance, when you buy an airplane ticket, you check the open seats in a given flight and the price. To achieve this, several small queries are served by the database in a short period of time, and the data involved in the answer is a small set of the data stored.
Online analytical processing (OLAP)
In the case of OLAP, you send very heavy queries that need to process a huge amount of data in comparison to the total size of the database. You send fewer queries to your system, but very heavy ones. In our airline example, if you were working in marketing you might want to know how many people between 20 and 35 years old traveled from New York to Madrid during 2014, grouped by fares and describing how far in advance they purchased their tickets.
So, we have different uses for the same set of data. In this scenario, some database vendors offer a one size fits all solution, suggesting that just one machine is able to address such different workloads at the same time.
IBM Hybrid Approach to Transactional processing and analytics
Through IBM DB2 Analytics Accelerator, IBM supports a different approach. This DB2 Analytics Accelerator, together with DB2 for z/OS, form a self-managing, hybrid workload-optimized database management system that runs each query workload in the most efficient way. With this approach, you avoid the headaches involved with the configuration of a database designed for OLTP that is also trying to serve OLAP workloads. This is the main reason why many data warehouses are difficult to manage, expensive to maintain and require many people to tune them—while still producing frustrated end-users.
IBM DB2 Analytics Accelerator turns DB2 for z/OS into a universal database management system, capable of handling both transactional and analytical workloads.
IBM DB2 Analytics Accelerator
IBM proposes the use of a hybrid environment where each query workload is executed in the most optimal environment for maximum speed, execution and cost efficiency. This hybrid infrastructure blends the best attributes of symmetric multiprocessing (SMP) leveraging DB2 for z/OS with the best attributes of the hardware-accelerated massively parallel processing (MPP) architecture delivered by Netezza technology.
The hybrid environment is enabled by the addition of the DB2 Analytics Accelerator, a high-performance appliance that integrates the z Systems infrastructure with PureData System for Analytics, powered by IBM Netezza technology.
And yes, it is only for DB2 for z/OS, at least for now.
This solution generally works as if you just added another access path that is specialized for processing analytic queries to your mainframe. Because it is just like an additional access path, query processing happens transparently so that users and applications can send the very same DB2 query requests to the system, unchanged.
The interface for the end-user does not change at all. And when I talk about an “additional access path”, what I really mean is that we add the DB2 Analytics Accelerator to complement DB2 for z/OS, which is built for transactional workloads. The Accelerator provides a cost-effective high-speed query engine to run complex analytics workload. Therefore, IBM DB2 Analytics Accelerator turns DB2 for z/OS into a universal database management system, capable of handling both transactional and analytical workloads.
As you can see, there are two different machines, each one solving different workload needs working as one system. As part of its unique design, the DB2 Analytics Accelerator includes breakthrough technologies to reroute complex, data-intensive queries to the integrated IBM PureData System for Analytics appliance. But the key point is that nothing has changed for the end user when compared to their traditional DB2 database, except that suddenly the analytical queries run much faster than before, with MIPS consumption decreasing. So, although you have two machines, they compound into a single system for the end users. And from the point of view of the administrators, there is no added complexity either.
DB2 Analytics Accelerator includes breakthrough technologies to reroute complex, data-intensive queries to the integrated IBM PureData System for Analytics appliance. But the key point is that nothing has changed for the end user when compared to their traditional DB2 database, except that suddenly the analytical queries run much faster than before . . .
In the words of one of our customers: “We are surprised how easy it is to manage such a system, it is really ‘plug-and-play’ and the results are awesome”.
This concludes my introduction. In following posts I’ll explain how DB2 Analytics Accelerator works, as well as real life experiences with it. To learn more, visit the DB2 Analytics Accelerator page on ibm.com.
Meanwhile, you can leave your comments, share your experience or join me on a conversation on Twitter.
See additional posts