Research Library

A Brief Guide to Scaling Oracle Databases

Big data has put a big squeeze on businesses, as data grows much faster than IT budgets: Gartner forecasted a worldwide overall IT spending increase of only 3.8% for 2014,1 while global information volumes rising over 59% each year.

Since companies need all their data to glean insights about their business, organizations are now looking for ways to handle exploding data volumes while reducing costs and maintaining performance.

These trends are overwhelming both databases and budgets, particularly when it comes to Oracle, which prevails as the most popular relational database  management system or RDBMS.3 Scaling Oracle solutions can be difficult and expensive. Managing large volumes and achieving high levels of concurrency often means expensive, scale-up hardware.

As such, many businesses with Oracle databases have began to investigate solutions that can affordably scale-out on commodity hardware. Scale-out solutions spread the data and the computation across a cluster of inexpensive  machines in parallel instead of centralizing data and computation on more expensive hardware. 

On the surface there are many alternative scale-out technologies, but it turns out that most are inadequate for current Oracle database users:

• NoSQL solutions (e.g., MongoDB, Cassandra), by design, lack SQL, joins, aggregations, and transactions, which will force major rewrites for any application currently using an Oracle database. 

• SQL-on-Hadoop solutions (e.g., Hive, Impala), designed only for ad-hoc analytics, are unable to support real-time operational applications, which often require transactional updates and a high concurrency of small reads and writes.

Thus, for organizations that are looking to scale affordably with a proven scale-out technology but still maintain full SQL support and RDBMS functionality, a Hadoop RDBMS is the obvious answer.

Harte Hanks, a global digital marketing services provider, selected Splice Machine, a Hadoop RDBMS, to replace its Oracle RAC databases and saw the following results:

• Queries become 3-7x faster

• Over 75% less expensive

• Well over 10x better price/performance

• Dramatically simplified scaling going forward

Harte Hanks was able to achieve these results without rewriting its IBM Unica campaign management software, Cognos business intelligence reports, Ab Initio ETL scripts, or Trillium data quality software.

Offered Free by: Splice Machine
See All Resources from: Splice Machine