If you’re like me, one of the most frustrating things about applications and databases is not knowing how long it is going to take to respond. A query that takes ½ a second one day and then 5 seconds the next, and 8 the day after that not only frustrates users, but makes more work for the DBAs. Why do I say this creates more work for the DBAs, because you can be sure that the users are calling and complaining about the varying response times, and the DBAs are monitoring the system to try and decipher what is going on.
You might think that this “variability” is the same for all databases, but that is not true. The data streaming architecture of the PureData System for Analytics pushes the processing to the data, meaning that it processes data at the speed of physics. The data is processed as soon as, and as fast it is streamed off of the disks. The PureData System for Analytics does not rely on large memory caches, indexes, or aggregates to perform well. That means that no matter how complex the SQL query, not matter what column or columns are accessed, the PureData System for Analytics provides consistent, predictable results.
What would you say if I told you I can tell you how long any query will take, based on the table it is accessing? Well that is exactly the case for the PureData System for Analytics. And, unlike databases that rely on indexes and aggregates for performance, if you change the query to:
- Retrieve another column or more columns
- Add new predicates
- Change the order of the predicates in the query
The PureData System for Analytics will continue to run with the same, consistent response times.
Ask you IBM sales team (and our competitors) how we do that, and we’ll even prove it to you on your site, with your data, and your queries.