Oracle Voice, 06/23/2014
By: Oracle Guest, Tom Haunert
To understand how in-memory technology translates into faster business processes, listen to Larry Ellison.
When Oracle ORCL -1.14%’s CEO introduced the new Oracle Database In-Memory option on June 10, he gave example after example of how this new technology accelerates the performance of everyday business applications: A cost management application performed 257 times faster. A financial analyzer app crunched numbers 1,300 times faster. A sales order analysis app was 1,700 times faster.
In one test case, a workload that ordinarily took 4 hours was accomplished in just 4 seconds—so fast that the technologists on hand weren’t quite sure anything had actually happened. “People say, ‘I don’t think it ran,’” said Ellison. In fact, it did.
How do you explain these eye-popping performance gains? Oracle Database In-Memory boosts performance through an architectural twist to the relational database model that has long been the norm in corporate data centers.
For years, relational databases have stored information in related tables, which are composed of rows and columns. Think of multiple spreadsheets, but with millions of rows of information and columns shared between different tables that connect the separate spreadsheet information.
Table data has traditionally been stored on disk and in computer system memory in row format—in a row store—because business transactions—for an employee, customer, invoice, manufactured part, call record, and so on–all alter table information at the row level. Queries against the database typically return information on a few table columns, but queries—especially analytic queries—may also need to look at all of the rows of table data to return a result.
Oracle Database In-Memory adds a column store to the traditional row store, an addition that significantly speeds up both analytic queries and complex transaction processing. Its dual-format approach simultaneously organizes data in both rows, for optimal transaction performance, and columns, for high speed analytics.
There’s much more to this advance than raw speed. Here are 12 things about Oracle Database In-Memory that are new and different.
1. It’s an option for Oracle Database 12c. Oracle Database In-Memory is not a standalone product, but rather an option for Oracle Database 12c Enterprise Edition. So Oracle Database 12c—introduced last year and now Oracle’s flagship relational database—now has an optional in-memory column store.
2. It’s actually more than in-memory. Oracle Database In-Memory puts a new column store into fast RAM (random access memory). The row store, meanwhile, continues to read from and write to memory, flash, and disk, and syncs information with the column store. The column store exists only in memory.
3. Database performance is very fast. When previewing the technology last year, Oracle stated that Oracle Database In-Memory could deliver query results 100 times faster, while doubling transaction processing rates. At the June 10 launch event, Oracle reaffirmed those numbers in customer test results.
4. It’s even faster than that. As mentioned above, Larry Ellison pointed to test scenarios where Oracle customers accelerated analytic queries by far more than 100 times using Oracle Database In-Memory. Oracle’s In-Memory Transportation Management app ran 1,030 times faster, and the Peoplesoft In-Memory Financial Analyzer executed 1,300 times faster.
5. Analytics get a big boost from the new column store. With Oracle Database In-Memory, all analytic queries run against the column store—not an index and not the row store. The column store is optimized to run analytic queries up to 100 times faster (or faster still, according to some customer tests).
6. Online transaction processing can run twice as fast. OLTP can run twice as fast with Oracle Database In-Memory column store because analytic indexes—which can speed known queries against the row store—no longer need to be created or updated for queries against the column store.
7. No changes are required to applications or queries. Applications and analytic queries continue to run against Oracle Database 12c. Oracle Database In-Memory changes the location of the query processing to the column store, but application and query code do not change.
8. Setup and maintenance are easy. Oracle Database In-Memory setup involves configuring the capacity of the column store, specifying what tables (or partitions—a database storage construct) to include in the column store, and deleting no-longer-needed analytic indexes. That’s it.
9. Fault tolerance is included. Oracle Database In-Memory is Oracle Database 12c; the in-memory column store is synchronized with the row store; and all of the reliability and disaster recovery features of Oracle Database 12c, including Oracle Real Application Clusters, protect information in the column store.
10. It’s ready for big data. For applications and queries that run against large databases with increasing data volumes, increased analytic index use, and long-running queries, Oracle Database In-Memory delivers real-time analytics. The size of the database is unlimited with Oracle Database In-Memory.
11. Many different queries are supported. If the business needs to ask an unlimited number of business questions about a constantly expanding set of business information, Oracle Database In-Memory can provide instant answers.
12. Dual-format architecture brings flexibility. Oracle Database In-Memory is for mission-critical databases and applications that do OLTP and business analytics. Separating transaction processing and reporting databases has been one way to optimize both types of operations, but at a cost of maintaining separate system hardware, software, and processes to sync the systems. With Oracle Database In-Memory, you can consolidate transaction and reporting databases, provide capacity for the database to scale up, do transaction processing and analytic queries through one cloud-hosted database instance, and provide high-speed transaction processing and instant answers in mobile applications.
Oracle Database In-Memory is scheduled to be available in July. For more on this new technology, check out the September/October 2014 issue of Oracle Magazine.
You can view a video replay of the Oracle Database In-Memory launch with Larry Ellison here.
And Larry Ellison’s Oracle Database In-Memory presentation is available here on Slideshare.com.
Tom Haunert is Editor in Chief of Oracle Magazine.