Horses for Courses
A road trip to the Kimberley region confirms the wisdom of selecting optimized systems for specialized workloads. Outback Australia is tough country, and the locals drive vehicles specifically designed and built for this terrain - predominantly Mitsubishis, Nissans and Toyotas. Off-road driving’s demands impel convergence in function and design; all these vehicles maintain good traction on uneven and unstable surfaces (low gearing, locking differential, long travel suspension, four wheel drive, heavily lugged tires) and their bodies’ entry and exit angles allow them through tight spots with minimal damage.
Last week, I learnt of a European communications company’s experience with web analytics. The company licensed Unica NetInsight and, given their enterprise database license, deployed the application on Oracle. Highly visited, public-facing web sites generate large volumes of data. The business users suffered a loss of confidence in their newly deployed application when forced to wait hours for reports and analyses. The business asked IT for help. The technical team partitioned their Oracle database, but tuning it brought little relief - the business still waited hours for their results.
Next, the IT team suggested upgrading the servers running Oracle, but could not guarantee that this would solve performance problems. IBM’s account team recommended that the company undertake a four-week proof-of-concept with Netezza, and the client agreed. Following a straight-forward data migration, the client’s IT team connected Unica to the Netezza database and ran a suite of reports selected by the business.
Consternation spread through the team: something must be wrong! Had the data load aborted? Reports that previously consumed hours of processing returned their results in seconds. A quick check showed all the data loaded in Netezza.
The technical team hastily convened a meeting with the business for the following day where they showed a live demonstration of Unica running on IBM Netezza. So confident were the team that they took ad-hoc requests from the packed room and ran reports and analyses never attempted on their old database system. Witnessing near-instant response times, one manager exclaimed: this is the application we always wanted.
The application hadn’t changed, but matching it with an appliance optimized for analytic workloads allowed it to fulfill its promise, and rather than disrupt the business, the database simply did its job and so became invisible.
Off-road driving creates different demands on a vehicle than cruising along a freeway and, while sharing basic design features with regular road cars, off-road vehicles are optimized for their workload. Putting knobbly tires on a road car cannot equip it for the difficult terrain encountered in remote regions.
Analytic and transactional applications create different workloads on their databases – adding partitions and adopting other tuning techniques cannot equip a database management system designed for transaction processing for the challenges of analytic processing. If, like the communications company, your analytic applications suffer poor performance, try running them on an appliance optimized for these workloads.