One of the most frequently asked questions by customers is:

Does SAP on Solaris x64 perform better or worse than on Linux or Windows ? The latest two-tier SAP-SD standard application benchmark results provide a clear answer to that question. Although I have to admit, that it's still not a 100% apples to apples comparison, you will be able to form your own opinion about this subject after looking at the facts. Below I added a comparison of benchmarks performed by HP using SAP ERP 6.0 (2005) non-Unicode with Oracle 10g RDBMS and by Sun using SAP ERP 6.0 (2005) Unicode with MaxDB 7.6 RDBMS.

Note:
It perfectly makes sense to really understand the impact of Unicode with respect to sizing. It may be important for your SAP landscape in case you are planning for a Unicode migration. With respect to benchmarking it is important to know, that in the near future SAP will exclusively publish Unicode benchmarks. Therefore you may be forced to compare 'older' non-Unicode results with current / future results based on Unicode. I highly recommend to make yourself familiar with this topic. Hopefully this posting provides a good start for you. You may check Understanding a SAP SD 2-tier benchmark result for more details.

While the impact of the database software is somewhat negligible (the two-tier SAP SD benchmark is much more application-centric than database centric), Unicode has a substantial impact on application server throughput:

• First, the SAP ABAP application server uses UTF-16 for data representation. Therefore you need way more memory to process the same amount of data. In other words: To achieve the same throughput (SAPS), you need significant more memory. Which then again requires more memory bandwidth ....
  According to SAP, you need 40 – 50 % more memory for the SAP application server. Please check SAP note 1139642 for details (https://service.sap.com/sap/support/notes/1139642 Note: User and password for SAP Service Marketplace required).

• Second, SAP ABAP application servers running Unicode require significantly more CPU power. This depends on the OS as well as the specific hardware. The same SAP note referenced above lists an additional CPU load of 10 – 30 %.

• Third, we don't look closer at the impact of Unicode on database performance due to the reason mentioned earlier: the two-tier SAP SD benchmark is very application-centric

Now you would expect, that two servers equipped with the exact same number and type of processors and the same amount of RAM would deliver similar results if running the exact same application. Furthermore, you would expect that running the two-tier SAP SD standard application benchmark using non-Unicode representation would deliver significant better results (more SAPS) than running the same benchmark suite using Unicode. Wait a minute, didn't we also talk about different Operating Systems ? Yes, indeed ! Choose Solaris as the OS for the Unicode benchmark and things change dramatically. Despite running Unicode, the Sun benchmark delivered 11.6 % higher throughput than a benchmark of HP using SuSe Linux Enterprise Server 10. Add the 10 – 30 % Overhead caused by Unicode and calculate your own 'Solaris Factor' !

Now look at the facts and convince yourself that there's no hitch to it:

For more details, see http://www.sap.com/benchmark .

You can see, the hardware of both systems is quite similar. Of course, there are some design differences: The Sun Fire X4600 M2 is a more compact design requiring less rack space (only 4 RU compared to 7 RU). There are also differences in the way the memory and the processors are connected to the mainboard. This can cause a slight advantage on one side, but it's unlikely that hardware is the only cause of such a considerable difference. Solaris is certainly the other important factor.

Now we could talk about a comparison of MS Windows Server and Solaris 10. To make a long story short: basically you will get to the same conclusion when comparing the SAP SD standard application benchmarks with:

ID 2008061:
Sun Fire X4600 M2, Solaris 10, SAP ERP 6.0 Unicode, MaxDB 7.6, 8 processors AMD Opteron 8360 SE @ 2,5 GHz meaning 32 cores and 32 threads, 128 GB RAM, 5800 SD users resp. 29670 SAPS

and

ID 2008026:
HP ProLiant DL785, Windows Server 2003 EE, SAP ERP 6.0 non-Unicode, SQL Server 2008, 8 processors AMD Opteron 8360 SE @ 2,5 GHz meaning 32 cores and 32 threads, 128 GB RAM, 5230 SD users resp. 26180 SAPS

By the way:
the difference between the two SAP SD standard application benchmarks on Solaris 10 are the type of processors being used. In benchmark 2008061 previous generation Quad-Core AMD Opteron 8360SE (codename Barcelona) were used while benchmark 2008070 was based on latest Quad-Core AMD Opteron 8384 technology (codename Shanghai). Look at the details and you will find a difference of nearly 35% - which is at least far more than I expected ! Good progress by AMD.


SAP and all SAP logos are trademarks or registered trademarks of SAP AG in Germany and in several other countries.