What shall and what shall not be used in database development in 2013
Though the current databases need new solutions, the well-known RDBMS is still active and working. Still there are places and cases where it shall not be used for your own sake. Even the people who are really fond of Oracle, RAC or any other system of this kind shall reconsider their urge to use their favorite ways in the following cases.
Do not use Oracle Text. It’s not usually used even by the Oracle lovers, since this database extension doesn’t develop actively. The other ways database developers tend to use in Oracle are either too primitive or too complicated; both options offer the bad kind of results.
However you can easily use Autonomy, Apache Solr or Hibernate Search for your task. This way you will get fast and handy results.
If you work with ATG Commerce, try to avoid this action by all means. The reason for this option to be turned off most of the time is the great scalability – database developer just does not need that many recommendations for his work. This feature is not usually needed at all, since the tasks people solve are really various, so you just do not need recommendations based on someone else’s actions.
High frequency trading
This is one of the first cases when people really tried to avoid it and even created the main NoSQL approach for it. Most people tend to use HFT for this matter with all its Riak or Gemfire benefits. Still some Oracle lovers try to manipulate with Times Ten with no actual use.
Cataloging of products
A lot of specialists in the field of database security can only imagine just what the hell is to create SQL queries for product data. Since sometimes good may have the same name but different components, it’s easy just to mess up.
The problem can be easily solved by Neo4j or any other graph database. Even if you have to opt for MongoDB, CouchBase 2.0 or SQL, do not choose the last one.
ACLs and users/groups
At the very beginning LDAP was created as the original NoSQL database a lot of database developers loved to work with. However, the level of database security out by some companies crashed the whole idea of the handy system, so a lot of people had to cheat and create the local database table.
All this led to IT specialists in the field of database development feeling the real hangover from the offered technology.
In order to see how complicated everything is, turn on the analysis log features and set the log saving for everything but ERROR. That is the best that can be done here, since the moment you ask for anything more complicated, you will get stuck.
Languages like PIG and applications like MapReduce offer much more freedom with low latency.
Repository of media
It’s fine while you work with metadata, still BLOBs will be total disaster. All the actions can be successfully performed in MongoDB, Couchbase or any other kind of document database.
It’s really hard to integrate email into any kind of RDBMS project. Even after the general optimization the main things do not experience any great changes. The only solution is to keep metadata off in your documental database, and the file system will be just fine.
Here the MongoDB will be the best solution ever. A lot of web developers prefer to opt for its Craigslist with adequate offering of content.
Forecasting of time and series
It can be offered in various forms like commodities, quants etc. Even the weather and sunspots will be here as well. You can do it nice and slow with the help of RDBMS, and the “slow” word is the meaningful here. If you want everything to be done fast, use the MapReduce-friendly store. Cassandra will go.
Sure thing you can go on using RDBMS for all these 10 cases, a lot of people do after all. Still if you want to save your time, try any of the offered ways. We truly hope you’ll be satisfied.