In this post I am going to go into a little deeper comparison of the PureData System for Analytics (the new generation of the IBM Netezza Appliances) and the Exadata Engineered System.
The PureData System for Analytics has a 10 year history of being expertly integrated so that it can be running in hours, not days, and does not need an army of DBAs to get it up and running and then keep it running. The unique hardware acceleration in the Field Programmable Gate Arrays (FPGAs) and the Asymmetric Massively Parallel Processing (AMPP) architecture allow the system to quickly get the results your users want, without the need for complex tuning, like indexes, aggregates, partition strategy, etc.
At Oracle Open World held in September 2012, a number of customers talked openly about their “migrations” to Oracle Exadata. Setting up their Exadata system in the data center took 10 days to two weeks, and “migrating” their Oracle database onto an Oracle Exadata engineered system took months to get up and running. Even Barry McGillin, Senior Development Manager at Oracle talked about a six step processes for migrating to Oracle Exadata in his session (CON8499) about Migrating to Exadata at Oracle Open World, September 2012.
These six steps include:
- Data Move
Contrast that with PureData System for Analytics (and IBM Netezza data warehouse appliance) customer Con-way Freight. Their initial set up took less than 48 hours, and within the next two days, data was loaded and users were already submitting queries. Four days after signing the purchase order, the new Business Intelligence (BI) project was made available to users. A mere three weeks from the purchase date, the platform was running full-scale production.
Oracle requires the DBA to design the data model to work well in the Oracle system. And this normally takes a number of iterations of trial and error, and/or the development of a multi-tiered data model, with the data being loaded into one physical model, and then transformed into star/snow-flake data models for reporting with different applications.
Instead of weeks of planning their data models, IBM PureData System for Analytics customers simply copy their table definitions from their existing tables, load the data, and start running their queries/reports/analytics. There is absolutely no need to create any aggregates or indexes. IBM PureData System for Analytics appliances are designed to deliver results almost immediately.
Previously I talked about the four data warehouse optimizations in Oracle Exadata and compared them to IBM Netezza/PureData System for Analytics. But there is another factor that is even more important. Analytics are driving a fundamental shift in the way companies do business. The IBM Institute for Business Value to conduct a survey of nearly 3,000 executives, managers and analysts working across more than 30 industries and 100 countries. Among our key findings:
- top-performing organizations use analytics five times more than lower performers.
- Overall, our study found widespread belief that analytics offers value.
- Half of our respondents said that improvement of information and analytics was a top priority in their organizations.
- And more than one in five said they were under intense or significant pressure to adopt advanced information and analytics approaches.
- Because of these there was a 57% increase in the respondents who stated that they believe that analytics creates a competitive advantage for them.
Executives have long been accustomed to a degree of imprecision and uncertainty when making decisions critical to their growth – and survival. For some companies their “best guess” was no longer good enough; hard facts were needed. These companies (the ones that use analytics) significantly outperformed their competitors by 220%!!!
But in Oracle Exadata, the analytic processing is done on the database servers, not in parallel across all of the storage servers. The Exadata storage servers might be able to do some rudimentary filtering of the data, but all of the data (or a sample in many cases) has to be moved onto the database servers to run the analytics, if Oracle even supports the analytic function that is being run. If not, then the data needs to be move again to the analytic server where the analytics will be run. And, whether the analytics is run on the database server or on an analytic server, the analytics will run against all of the data, with little to no parallelism.
In PureData System for Analytics the analytic processing is also pushed down to the data, and operates on the massively parallel system, or a parsed/reduced set of the data as fast as it streams off the disks. The IBM PureData System for Analytics includes over 200 built in complex analytic functions that have been optimized to run on this platform. The next closest data warehouse vendor has less than 50 built in analytic functions.
This means that with PureData System for Analytics you can run the analytics you need to run without constraints. And you can run it on your entire data set, not just a sample, giving far more accurate results.
The PureData System for Analytics hardware accelerated MPP architecture allows SUNY Buffalo to run quintillions of operations needed to model complex algorithms in their research for the cure for Multiple Sclerosis in minutes rather than the hours it took them before.
Now is the time to evaluate the PureData System for Analytics, but be prepared to do more with less, faster than ever before!
 Con-way Freight Case Study retrieved 10/15/2011 from http://www.google.com/url?sa=t&rct=j&q=con-way netezza&source=web&cd=5&ved=0CD4QFjAE&url=http%3A%2F%2Fwww.netezza.com%2Fdocuments%2Fcon-way-freight-case-study.pdf&ei=Xwy4ToTMDaOo2wWKjL3MDQ&usg=AFQjCNEgDL1PIvwHanu8uGh1Um8NTD6O4Q&cad=rja No two installations are identical, but the databases were of similar size and scope based on the available information
 Based on vendor’s documentation for the current version of their system/software