At an event at Oracle’s headquarters, Oracle CEO Larry Ellison showcased the new Oracle Database In-Memory, an additional option to the Oracle Database 12c. Oracle Database In-Memory is scheduled for release in July, said Ellison.
The real-time memory optimized capability was previewed at Oracle OpenWorld 2013. The actual launch represents the first maintenance release following the release of Oracle Database 12c in June 2013 and officially is designated as 220.127.116.11. “Maintenance releases are generally about bug fixes. This one is a little bit unusual in that it has this major new feature – the Oracle Database In-Memory option,” explained Tim Shetler, Oracle vice president of product management, in an interview before the launch event in which he highlighted the key elements of the release.
The Oracle Database In-Memory is the underpinning of finally making the real-time enterprise a reality, said Shetler. “We have built this in-memory option to be completely transparent to the huge ecosystem of Oracle Database customers and applications that are already written for the Oracle Database.”
No Application Changes
The key is not just performance but the ability to adopt the new technology within an existing environment and existing applications without having to make any changes to those applications, said Shetler. “The complete transparency is in contrast to all the other in-memory database options that exist out there today. The other thing we have designed this to do is run this real time capability, this in-memory analytics if you will, against the production transactional database. The ability to combine traditional transactional workloads and analyze the data in real time is a hallmark of the way we have implemented the in-memory option.”
The traditional way of doing analytics or reporting involves taking a copy of the production database and getting it off the server so you are not interfering with transactions, and then running reports and queries against that, but that takes time and, by definition, the data starts to become stale as soon as you have taken a copy of it, said Shetler. “Now we are letting people simply work with one database and do all of those workloads, all their transactions and all the analytics against the same database.
Dual Format Architecture
To enable this, said Shetler, Oracle Database In-Memory provides a dual-format in-memory architecture that combines the best of row format and column format to simultaneously deliver fast analytics and efficient OLTP. Oracle Database In-Memory allows any existing Oracle Database-compatible application to automatically take advantage of columnar in-memory processing, without additional programming or application changes.
“We have extended the optimizer to know whether the data that is being requested is in this new in-memory columnar store and if it is, it can transparently direct the request to that. But if it is a request to say, enter a new row, it will direct that request to the traditional row store.”
No Database Size Limits
The other aspect of the Oracle Database In-Memory that is unique, said Shetler, is the ability to run with any database of any size, said Shetler. “Some of the in-memory implementations require that the entire database fit into memory, and obviously, databases – particularly data warehouses – can be enormously large today – hundreds of terabytes or even petabytes in size. You can’t even buy a server or a cluster of servers that can hold a petabyte of data in memory today and you certainly wouldn’t want to pay for all that memory. Moreover, the truth is that in a database, the majority of the data is not really active. It is cold data and it should be sitting on cheap storage like disk.”
With the way Oracle has implemented the in-memory option, said Shetler, the customer’s data can be anywhere - it doesn’t have to be in the in-memory store. “You can use our clustering technologies to spread the data across multiple computers in a cluster; you can use something like the Exadata Database Machine to have portions of the data in memory, portions on flash, and portions on disk, and we will transparently find it all and pull it all together and do a query for you. We don’t tell you that it all only applies to your smaller databases. We tell you that the in-memory option is something that you can apply to any database in your environment – no matter how big it is.”
According to the company, Oracle Database In-Memory demonstrated from 100x to more than 1000x speed-up for enterprise application modules, including Oracle E-Business Suite, Oracle’s JD Edwards, Oracle’s PeopleSoft, Oracle’s Siebel, and Oracle Fusion Applications.
“We have gone to great lengths to make easy adoption and transparency a hallmark of what we are doing and the major differentiator within the In-Memory Database environment today,” said Shetler, noting that Oracle has included over 800 people in testing the in-memory database option so far.
Software and Hardware Engineered for Real-Time
The Oracle Database is now over 30 years old, said Shetler, and the in-memory option inherits all the previous development including security, availability, scalability, manageability, and all the work that has gone into making sure that transactions are not lost and data is not corrupted.
In-Memory and Engineered Systems
Oracle Engineered Systems, including Oracle Exadata Database Machine and Oracle SuperCluster, are optimized for Oracle Database In-Memory, featuring large memory capacity, extreme performance, and high availability while tiering less active data to flash and disk to deliver outstanding cost effectiveness. In-Memory fault tolerance on Oracle Engineered Systems optionally duplicates in-memory data across nodes enabling queries to instantly use a copy of in-memory data if a server fails. New Direct-to-Wire Infiniband accelerates scale-out for in-memory.
According to Oracle, the M6-32 Big Memory Machine is the most powerful scale-up platform for Oracle Database In-Memory providing up to 32 terabytes of DRAM memory and 3 terabytes/sec of memory bandwidth for maximum in-memory performance.
Support for In-Memory Option on Non-Oracle Platforms
In addition, the in-memory option is going to be available wherever the Oracle Database is available – not only Oracle systems, but also on any non-Oracle platform where the Oracle Database is available including commodity platforms, Shetler said.