Chasing the Oracle Exadata Pot of Gold
The contest that never was
Oracle is once again advertising a contest where Oracle claims that if an Exadata data warehouse system is not “5x faster” than a Power 795 data warehouse system, you win 10 million. Once again, the rules are clearly defined in a way that Oracle can never lose the contest that never was. My favorite: Oracle “reserves the right, at its sole discretion, to disqualify any entrant”.
Big purse “contests” generate media attention but don’t help anyone get any real work done. From my experience working with hundreds of customers who switched from Oracle to IBM, business users are interested in getting insight fast enough to have a big impact on business decisions. IT departments want help dealing with high database maintenance costs and on-going tuning challenges. And the answer to these challenges is more than just fake contests hyping Exadata.
Oracle claims that Exadata is optimized simultaneously for transactional databases and analytic workloads. Marketing gimmicks and $10 million prizes still can’t obscure the truth. According to Kevin Clossen, an ex-Exadata evangelist, “There is no value add for Oracle Database on Exadata in the OLTP/ERP use case. Full stop. OLTP/ERP does not offload processing to storage”. And by the way, Exadata offers far less flexibility than a general purpose system such as a Power System without really delivering appliance-like simplicity either. What it does though is restrict client choice with a one-size-fits-all system.
For real users with real business needs who focus on getting real work done, IBM takes a workload-based approach to system optimization, resulting in greater choice, flexibility and value to the client. I want to look at a few typical data-intensive workloads. Let’s start with analytics.
IBM Netezza is a data warehouse appliance specifically built to make complex analytics faster, simpler, and more accessible. It arrives with database, server and storage integrated into a single, easy to manage system. It costs a fraction of Orale Exadata1 and it’s up and running in 24 hours. The Netezza appliance requires little on-going maintenance and delivers stunningly fast responses to complex analytic queries that can span entire data warehouses with billons of records. Of course, the ultimate test before making a decision is a proof of concept. If you are looking to speed up your Oracle data warehouse, Netezza is likely to be the answer, not Exadata. We do a lot of test drives at client sites, with client data, and four out of five companies that try Netezza end up buying it.
Does this mean we always suggest Netezza to everyone struggling with data warehouse performance? No. One size does not fit all. There are analytic workloads with different system requirements and client environments with specific needs, where we would recommend a better fit.
The IBM Smart Analytics System is built for high throughput of operational analytic queries. Think of telesales reps in a large Call Center using an application that pulls information from the data warehouse every time a client calls. None of the queries are likely to be touching all the records but you have to be able to respond to everyone in near real time, before the client hangs up the phone. This means that you may end up with thousands of concurrent users running analytic queries against your operational data warehouse. That is what the Smart Analytics System is really good at. It arrives pre-integrated with compute nodes, storage, data warehouse, advanced analytics and business intelligence. All in one, with a single part number. You get a choice of platforms: x86, Power, or System z, to best fit your environment. You can choose storage and capability modules too. Terabyte pricing is available and you could end up spending up to 43% less over three years compared to Oracle Exadata2.
Transactional database workloads
OLTP / transaction processing is actually what most Oracle database clients do with the Oracle Database. When trying to reduce your database software maintenance cost, it is not at all clear how Exadata would help. There are alternatives, however, that may very well do so, as I have witnessed many a times with companies moving from Oracle databases and/or hardware. Consider DB2 on Power Systems.
DB2 on Power can deliver up to 3x the performance per core than Oracle Database on SPARC. That is based on results on two different benchmarks3. ‘Per core’ is the key word here. This means you may end up needing fewer cores for the work you are doing, which translates to using fewer software licenses. Speaking of which, DB2 software can be as low as 1/3 the cost of Oracle Database4. Not to mention virtualized environments where with IBM software you only pay for cores you use, which is not always the case with Oracle software.
Of course, immediately the question of database migration comes up. With PL/SQL compatibility now baked into DB2, we have seen clients take an application based on Oracle 9i (or higher) and stand it up on DB2 in two weeks. Moreover, in SAP environments you don’t even have to think about compatibility. DB2 is uniquely optimized for running SAP applications. SAP themselves switched their internal application production environment to DB2 and their database performance improved by 40 percent5. And that is a real world result, not just a benchmark test. We have seen other organizations reduce their applicable infrastructure costs in support of SAP solutions by as much as 25 to 50 percent5 when moving to DB2.
Good old fashioned proof of concept
You do have choice when trying to speed up Oracle data warehouses and reduce rampant database software expenses. Rather than jumping over a rainbow chasing the Exadata pot of gold, simply do a proof of concept with IBM. Then you set the rules, not Oracle, and you get to decide which solution meets your needs most cost effectively. Try IBMs Stop Think and Save page for more information.
1 Cost comparison is based on publicly available information as of 4/17/2012 for an Oracle Exadata X2-2 HP Full Rack and a full rack of Netezza 1000. Both include Y1 maint/support. The cost to acquire Netezza can be nearly as low as 1/6 of Exadata if a client is acquiring new Oracle database licenses and nearly as low as 1/2 if using existing Oracle database licenses.
2 Cost comparison is based on International Technology Group whitepaper “Cost/Benefit Case for Enterprise Warehouse Solutions”, June 2011. This document was developed with IBM funding.
3 Results on TPC-C: www.tpc.org (http://www.tpc.org) as of 4/3/2012 [IBM Power 780 (3 x 64 C)(24 Ch/192 C/768 Th); 10,366,254 tpmC; $1.38/tpmC; avail 10/13/10 v. Oracle SPARC SuperCluster w/T3-4 Servers (27 x 64 C)(108 Ch/1728 C/13824 Th); 30,249,688 tpmC; $1.01/tpmC; avail 6/1/11]. TPC-C is a trademark of Transaction Performance Processing Council.
Results on two-tier SAP Sales and Distribution (SD) standard application benchmark, as of April 3, 2012: IBM Power 795 (32 processors, 256 cores, 1,024 threads) achieved 126,063 SAP SD benchmark users running the SAP enhancement package 4 for the SAP ERP application 6.0, AIX 7.1 and DB2 9.7; certification number 2010046 v. Oracle SPARC Enterprise Server M9000 (64 processors, 256 cores, 512 threads), which achieved 39,100 SAP SD benchmark users running SAP ERP 6.0, Solaris 10, Oracle 10g; certification number 2008042. SAP and all SAP logos are trademarks or registered trademarks of SAP AG in Germany and several other countries.
4 Cost comparison is based on publicly available U.S. info on 04/03/2011 for IBM DB2 Advanced Enterprise Edition + Oracle software w/comparable capabilities. IBM: 100 Processor Value Units. Oracle: assumes 1.0 processor multiplier. Both incl. Y1 maint/support. Cost comparison is NOT based on the specific benchmarks listed on this page.
5 SAP IT selects IBM DB2 as strategic database platform for internal business systems
Based on actual customer case studies. All client examples cited or described are presented as illustrations of the manner in which some clients have used IBM products and the results they may have achieved. Actual environmental costs and performance characteristics will vary depending on individual client configurations and conditions. Contact IBM to see what IBM can do for you.