Oracle Database In-Memory

Oracle Database

Trust Verified: Oracle Database Outperforms SAP HANA

By John Soat

 

Yogi Berra once said, "If you don't know where you’re going, you might wind up someplace else." In the Yankees catcher’s inimitable way, he was speaking to the necessity of being clear-eyed and laser-focused in pursuit of your goals.

That’s true when it comes to technology and business strategy: Stay laser-focused on your best opportunities and clear-eyed about the most effective tools to make them happen. And that’s especially true for SAP customers, as the enterprise software company points them in a new database direction, HANA, and away from their time-tested database platform, Oracle.

Now, benchmark tests demonstrate decisively that Oracle's most recent database technology, Oracle Database In-Memory, surpasses SAP's HANA in-memory database in performance and scalability.

The tests provide further evidence that Oracle offers the best database technology for the enterprise, including SAP customers. But before delving into the benchmark tests, first some background on the 27-year-old Oracle-SAP relationship and the emergence of in-memory databases.

A Most Successful Product

The most recent version of Oracle’s flagship technology—Oracle Database 12c, introduced in 2013—boasts several compelling add-on features, such as a multitenant architecture. By far the most popular new feature is Oracle Database In-Memory.

Oracle Database In-Memory is intended to speed up database performance, especially for data-intensive applications related to business analytics. It does so by placing data in memory in a new columnar format known as the In-Memory column store, an optimal arrangement for analytic queries. The new columnar format doesn’t replace the existing row format; instead, it’s a unique dual-format approach that allows data to reside in memory in both formats, row and column, which ends up boosting both transaction processing and analytics operations.

It’s an approach that’s proved appealing. “[Oracle Database] In-Memory has been far and away the most successful new product offering since RAC,” says Andy Mendelsohn, Oracle executive vice president for database server technologies. Oracle Real Application Clusters, introduced in 2001, is widely used for high-availability clustering.

The list of companies that have signed up for Oracle Database In-Memory represents just about every industry, from telecom to insurance to science to government. "The customers are very excited about this technology—they’re using it and they’re seeing spectacular results," Mendelsohn says.

One familiar telecom company has had Oracle Database In-Memory in production since October 2014. The implementation was quick (fewer than three months) and relatively effortless. More important, the company experienced a significant increase in performance—from 400 minutes to 10 seconds—in generating data analytics using SAP BusinessObjects.

The Original

Oracle introduced the first commercial relational database in 1979, and today Oracle Database is the world’s leading operational data management system. Oracle’s roster of partners, including those that base their software on Oracle Database, is long and familiar.

One partnership in particular, with enterprise software developer SAP, has been especially fruitful, dating to 1988. More than two-thirds of SAP’s midsize-to-large customers run their applications on Oracle Database. It’s multiplatform—companies can run SAP applications with Oracle Database on the same code base on Linux, Unix, and Windows—and it powers SAP’s most advanced software, including the SAP Business Suite, SAP NetWeaver platform and toolset, and SAP Business Warehouse.

Andy Mendelsohn

 Customers are very excited about this technology—they’re using it and they’re seeing spectacular results.  

—Andy Mendelsohn, Oracle Vice President for Database Server Technologies

 

SAP is well aware of the importance of Oracle technology to its customers. This is why SAP "certifies" releases of Oracle Database for use with its business applications. For instance, SAP certified Oracle Database 12c in March 2015, even while many SAP customers still use the (certified) prior release, Oracle Database 11g. SAP certified Oracle Database In-Memory on June 30, 2015.

The association between Oracle Database and SAP software is both deep—in terms of Oracle’s ability to run SAP applications at peak capacity—and wide, with Oracle supporting almost all of the company’s offerings. For example, several cloud applications SAP gained through acquisitions, including Ariba (e-commerce), Concur (travel management), and SuccessFactors (human resources), run on Oracle databases in SAP’s cloud.

No Confusion

In 2010, SAP introduced HANA, its in-memory database management system. Since then, SAP has touted HANA as the database of the future for its enterprise applications—all while still certifying Oracle Database for its customers and licensing it for itself. The fact that HANA still does not power all of SAP’s applications, including some of its cloud services, is revealing. An SAP executive even admitted that the company’s HANA strategy has "confused" its customers.

There’s no confusion for SAP customers about Oracle Database. Benefiting from decades of innovation, Oracle Database represents the industry’s gold standard in terms of built-in availability and data protection characteristics, enabling customers to maximize application uptime during both planned maintenance activities and unexpected failures. In contrast, SAP HANA lacks many of the fundamental high-availability capabilities needed to run today’s always-on, 24/7 enterprises.

Oracle Database isn’t lacking—it’s gaining. Oracle is the proven leader in transaction processing performance as measured by SAP's own standard application benchmarks. Now, Oracle Database In-Memory enables customers to perform real-time analytics on live transactional data.

Conversely, SAP HANA is optimized for analytic performance while its ability to deliver competitive transaction-processing performance is unproven. To that end, ask SAP why it has never published a single result demonstrating HANA’s transaction-processing capability using its own standard application benchmarks.

Head to Head

Yet SAP has sought a way to establish HANA as a viable alternative to the much more widely used Oracle Database. In 2012 SAP created its BW-EML (Business Warehouse-Enhanced Mixed Load) benchmark test—one of the many standard application benchmarks SAP has developed over the years—to measure typical data warehouse workloads in two areas: real-time reports and ad hoc queries.

Such tests are a speed contest of sorts, establishing database performance characteristics by running the software on a published configuration of supporting technologies—hardware, including CPU and storage; software, such as a particular application and database; and networking.

The BW-EML benchmark employs SAP Business Warehouse as the data warehouse application. It simulates a large number of users running online reports for a 60-minute period, and new data is loaded into the database at 5-minute intervals, simulating real-time updates. The object is to measure the total number of query navigations or decision steps completed in an hour’s time.

Among the benchmark’s variables is the scale factor: BW-EML can be run with databases storing from 500 million to 10 billion records. Results based on 1 billion records likely are appropriate to most enterprise use cases.

In October 2014, SAP released results for the BW-EML benchmark running its HANA database. SAP engineers ran the benchmark using a Dell server with two 18-core Intel Xeon E5-2699 v3 processors and 512 GB of memory. At the 1 billion records scale, SAP HANA completed 148,680 "navigation steps" in the given 60-minute period.

Seven months later, SAP released new BW-EML benchmark results for a single database server. SAP engineers again used a Dell server, but this time it had four 18-core Intel Xeon E7-8890 processors and 1536 GB main memory. At the same 1 billion records scale, but with this new configuration, SAP HANA achieved 320,940 navigation steps in the given hour.

Linear Scale

Earlier this year Oracle engineers turned to the BW-EML benchmark test. They employed a single database server in an Oracle Exadata Database Machine X5-2, which features up to eight servers, and used the same Intel processors and memory SAP used in its October 2014 results. Of that 512 GB of memory, 275 GB was allocated to the In-Memory column store. At the 1 billion records scale, Oracle Database 12c completed 300,749 "navigation steps" in the hour period—equivalent to 84 queries per second.

So, running on identical Intel processors and memory, Oracle Database In-Memory outperformed SAP HANA by a factor of 2x (300,749 navigation steps vs. 148,680). Additionally, Oracle’s technology reached a similar throughput to SAP’s second benchmark result (300,749 vs. 320,940)—even though SAP’s engineers used twice the number of cores and three times the amount of memory to achieve that result.

Oracle engineers also conducted the test using two database servers on Oracle Exadata X5-2, then four servers, then the full eight. Oracle Database In-Memory scaled almost linearly from configuration to configuration, finally achieving more than 2.3 million navigation steps per hour, or 650 queries per second, running on eight database servers.

In terms of scaling, SAP released BW-EML results in May 2015 at the one billion records scale for a seven-server configuration. And even though SAP employed almost twice the number of cores and double the memory in its seven-server configuration as Oracle did in its eight-server test, SAP HANA fell short of Oracle 12c In-Memory in completing navigation steps in the hour’s time (1,992,570 vs. 2,346,594).

Important Lessons

The key point to glean from the BW-EML benchmark results is Oracle’s cost/performance advantage: Oracle Database In-Memory outperformed SAP HANA by a factor of two on the same Intel processor. Another important takeaway is the total cost of ownership benefits related to scaling systems, demonstrated by Oracle Database 12c’s ability to increase performance linearly as hardware capacity increased.

It’s also important to understand that those cost/performance/capability advantages translate to real-world business benefits.

A well-established European insurance firm is upgrading its aging systems by standardizing on Oracle Database In-Memory. The insurer was impressed with the sizable performance improvements it witnessed when it tested the In-Memory column store feature using a customer information database, an executive data warehouse, and its 20-year-old risk management system—all without changing or fine-tuning those systems.

It's a pragmatic example of how the Oracle Database In-Memory column store increases database performance in analytic applications and in transaction processing operations. "We can give you both, and that’s the unique thing," Mendelsohn says.

Take to Heart

It’s also a lesson in practical database superiority that SAP customers should take to heart when considering a possible move to a new database platform such as HANA. As described above, Oracle Database is the proven leader in transaction processing performance and reliability. Now we also have shown that Oracle is the leader in analytic performance, as demonstrated by SAP’s own benchmark. Plus, it’s a non-trivial task to swap out databases that have been tuned for specific applications in particular environments. It requires significant investment to rewrite customized software and retrain personnel on unfamiliar technology. Not to mention big bucks in system integration and consulting fees.

Finally, the ability to solve real-world problems is the most important criterion in deciding on technology. "At the end of the day, benchmarks are not very interesting to customers," Mendelsohn says. "What they really want to see are business proof points."

Oracle Database, specifically Database In-Memory, has plenty of those proof points.

(Note: SAP has neither certified nor endorsed Oracle’s BW-EML benchmark results.)


Date of Submission
to SAP for
Review
Date of Measurement Technology Partner Number of records Ad-Hoc Navigation Steps/Hour Operating System - Release Database Server Database Release SAP NetWeaver Release Number & Type of
Database/Application Servers
 
 02/27/2015
 04/20/2015*
02/23/2015 Oracle 1,000,000,000 300749 Oracle Enterprise
Linux 6
Oracle 12c SAP NetWeaver 7.40
1 database server: Oracle Exadata Database Machine X5-2 Full Rack EF (1 compute node up) 2 processor / 36 cores / 72 threads, Intel Xeon Processor E5-2699 v3 2.30 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 45 MB L3 cache per processor, 512 GB main memory
4 application server: Oracle Exalogic Elastic Cloud X4-2 Exalogic Elastic Cloud Quarter Rack (4 compute nodes up) 4 x 2 processor / 4 x 24 cores / 4 x 48 threads, Intel Xeon Processor E5-2697, 2.70 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 30 MB L3 cache per processor, 4 x 256 GB main memory
 
 02/27/2015
 04/20/2015*
02/22/2015 Oracle 1,000,000,000 602378 Oracle Enterprise
Linux 6
Oracle 12c SAP NetWeaver 7.40
2 database server: Oracle Exadata Database Machine X5-2 Full Rack EF (2 compute node up) 2 x 2 processor / 2 x 36 cores / 2 x 72 threads, Intel Xeon Processor E5-2699 v3 2.30 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 45 MB L3 cache per processor, 512 GB main memory
8 application server: Oracle Exalogic Elastic Cloud X4-2 Exalogic Elastic Cloud Quarter Rack (8 compute nodes up) 8 x 2 processor / 8 x 24 cores / 8 x 48 threads, Intel Xeon Processor E5-2697, 2.70 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 30 MB L3 cache per processor, 8 x 256 GB main memory
 
 03/06/2015
 04/20/2015*
03/05/2015 Oracle 1,000,000,000 1209839 Oracle Enterprise
Linux 6
Enterprise Linux 6 Oracle 12c SAP NetWeaver 7.40
4 database server: Oracle Exadata Database Machine X5-2 Full Rack EF (4 compute node up) 4 x 2 processor / 4 x 36 cores / 4 x 72 threads, Intel Xeon Processor E5-2699 v3 2.30 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 45 MB L3 cache per processor, 512 GB main memory
24 application server: Oracle Exalogic Elastic Cloud X4-2 Exalogic Elastic Cloud Full Rack (24 compute nodes up) 4 x 2 processor / 4 x 24 cores / 4 x 48 threads, Intel Xeon Processor E5-2697, 2.70 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 30 MB L3 cache per processor, 24 x 256 GB main memory
 
 09/02/2015 03/12/2015 Oracle 1,000,000,000 2346594 Oracle Enterprise
Linux 6
Oracle 12c SAP NetWeaver 7.40
8 database server: Oracle Exadata Database Machine X5-2 Full Rack EF (8 compute node up) 8 x 2 processor / 8 x 36 cores / 8 x 72 threads, Intel Xeon Processor E5-2699 v3 2.30 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 45 MB L3 cache per processor, 512 GB main memory
32 application server: Oracle Exalogic Elastic Cloud X4-2 Exalogic Elastic Cloud Full and Eighth Rack (32 compute nodes up) 32 x 2 processor / 32 x 24 cores / 32 x 48 threads, Intel Xeon Processor E5-2697, 2.70 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 30 MB L3 cache per processor, 32 x 256 GB main memory
 *Note: We resubmitted results at SAP's request for a second review 4/20/2015.

 

Date of Certifcation Technology
Partner
Number of records Ad-Hoc
Navigation
Steps/Hour
Operating System - Release Database Server Database Release SAP
NetWeaver Release
Number & Type of
Database/Application Servers
Certification
Number
 
10/20/2014 Dell 1,000,000,000 148,680 SuSE Linux
Enterprise Server 11
SAP HANA 1.0 SAP
NetWeaver
7.31
1 database server: Dell PowerEdge R730, 2 processors / 36 cores / 72 threads, Intel Xeon Processor E5-2699 v3, 2.30 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 45 MB L3 cache per processor, 512 GB main memory
2 application servers: Dell PowerEdge R910, 4 processors / 40 cores / 80 threads, Intel Xeon Processor E7-4870, 2.40 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 30 MB L3 cache per processor, 512 GB main memory
2014038
 
05/05/2015 Dell 1,000,000,000 320,940 SuSE Linux
Enterprise Server 11
SAP HANA 1.0 SAP
NetWeaver
7.31
1 database server: Dell PowerEdge R930, 4 processors / 72 cores / 144 threads, Intel Xeon Processor E7-8890 v3, 2.50 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 45 MB L3 cache per processor, 1536 GB main memory
2 application servers: Dell PowerEdge R910, 4 processor / 40 cores / 80 threads, Intel Xeon Processor E7-4870, 2.40 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 30 MB L3 cache per processor, 512 GB main memory
2015015
 
05/05/2015 Dell 1,000,000,000 1,992,570 SuSE Linux
Enterprise Server 11
SAP HANA 1.0 SAP
NetWeaver
7.31
7 x database servers: Lenovo x3850 X6 , 4 processor / 72 cores / 144 threads, Intel Xeon Processor E7-8890 v3, 2.50 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 45 MB L3 cache per processor, 1024 GB main memory
15 x IBM Flex System x880 X6 Compute Node, 2 processor / 30 cores / 60 threads, Intel Xeon Processor E7-8890 v2, 2.80 GHz, 64 KB L1 cache and 256 KB L2 cache per core, 37.5 MB L3 cache per processor, 256 GB main memory
2015011

 


Get Started

Estamos aquí para ayudarle

Contrate a un experto en ventas

Prueba

Pruebe Oracle Cloud