SAN FRANCISCO: Oracle has added in-memory processing to its 12c database, aimed at speeding up queries by 100 times compared to the current model.
Oracle chief executive Larry Ellison revealed the update on Sunday evening, during his opening keynote at the OpenWorld show in San Francisco.
The in-memory option for Oracle Database 12c is designed to ramp up the speeds of data queries – and will also give Oracle a new weapon in the fight against SAP’s rival HANA in-memory system.
“When you put data in memory, one of the reasons you do that is to make the system go faster,” Ellison said. “It will make queries go faster, 100 times faster. You can load the same data into the identical machines, and it’s 100 times faster, you get results at the speed of thought.”
Ellison was keen to allay concerns that these faster query times would have a negative impact on transactions.
“We didn’t want to make transactions go slower with adding and changing data in the database. We figured out a way to speed up query processing and at least double your transaction processing rates,” he said.
In traditional databases, data is stored in rows, for example a row of sales orders, Ellison explained. These types of row format databases were designed to operate at high speeds when processing a few rows that each contain lots of columns. More recently, a new format was proposed to store data in columns rather than rows to speed up query processing.
Oracle plans to store the data in both formats simultaneously, according to Ellison, so transactions run faster in the row format and analytics run faster in column format.
“We can process data at ungodly speeds,” Ellison claimed. As evidence of this, Oracle demoed the technology, showing seven billion rows could be queried per second via in-memory compared to five million rows per second in a traditional database.
The new approach also allows database administrators to speed up their workloads by removing the requirement for analytics indexes.
“If you create a table in Oracle today, you create the table but also decide which columns of the table you’ll create indexes for,” Ellison explained. “We’re replacing the analytics indexes with the in-memory option. Let's get rid of analytic indexes and replace them with the column store.”
Ellison added that firms can choose to have just part of the database for in-memory querying. “Hot data can be in DRAM, you can have some in flash, some on disk,” he noted. “Data automatically migrates from disk into flash into DRAM based on your access patterns. You only have to pay by capacity at the cost of disk.”
Firms wanting to take advantage of this new in-memory option can do so straightaway, according to Ellison, with no need for changes to functions, no loading or reloading of data, and no data migration. Costs were not disclosed.
And for those firms keen to rush out and invest in new hardware to take advantage of this new in-memory option, Ellison took the wraps off the M6-32, dubbed the Big Memory Machine. According to Ellison, the M6-32 has twice the memory, can process data much faster and costs less than a third of IBM’s biggest comparable machine, making it ideal for in-memory databases.
Comcast's £29.7bn winning bid more than twice the £13.7bn Rupert Murdoch valued Sky at just eight years ago
A nuclear strike has been considered, but Bruce Willis is nowhere in sight
Spray-on antenna could enable seamless integration of antennas with everyday objects
Parker Solar Probe, TESS and GOLD missions will deliver exciting data, claims NASA