Tuesday, July 10, 2007

Why Sun Fire T2000 and Sun StorageTek 2540 for PostgreSQL Benchmark?

Based on the "First Real Benchmark for PostgreSQL"  publication, there was lot of discussion on Slashdot.org . One question was regarding selection of the server and storage which one claimed was "favorable" to PostgreSQL.  You bet!! But the same hardware is also favorable to other databases also. Try it out using the "Try and Buy" program of Sun.

But many points of the selection of the server and the storage did not get through and hence this blog entry. In this entry I will just cover few of the reasons of why there was change in configuration primarily when compared to the earlier MySQL/Glassfish result.
The intention was to provide a better price/performance number using an Open Source Database.
 

To drive the point home, I would like you to look at two Bill of Materials: One for MySQL/Glassfish and the other for the PostgreSQL/Glassfish configuration.  You will notice that the first change that the storage was replaced with the latest Sun StorageTek 2540. Why? Notice that the price of the storage in the MySQL/Glassfish configuration was about $23K and the new StorageTek 2540 now costs less than  $20K (with card). The 15K rpm drives of the new StorageTek array is huge added bonus over the 10k rpm drives of the old one. So is the added 2X capacity of storage all at a lower price. That selection was no brainer.

Why Sun Fire T2000? Specially that seems to indicate a  rise of cost compared to the earlier DB config.  The primary reason is to break the FUD  that competitors put forward that T2000 is not a good platform for OLTP databases. Now with an Open Source Database validation, it is easier for us to drive the point home that not only commercial databases like Oracle, DB2 perform well on Sun Fire T2000 but also Open Source Databases like PostgreSQL also perform quite well on Sun Fire T2000 (aka Sun SPARC Enterprise T2000).

Also one  might have noticed that price of EnterpriseDB premium support for PostgreSQL is based on per socket and not per core. When you take that into consideration you start seeing the savings involved already with Sun Fire T2000 because of the fact that Sun Fire T2000 has only ONE socket.

Plus the added advantage of IO throughput bandwidth and future growth in load and ability to do LDOMs with various operating systems versions of Solaris 10, Solaris Express, BSD and Ubuntu Linux gives the flexibility  to the enterprise for practically free of cost.


It will still be hard to convince all, but thats why we also sell Opteron and soon Intel based servers. You have a choice!!

 

Disclosure Statement:
Sun Fire X4200 (6 chip, 12 core)  - 778.14 SPECjAppServer2004 JOPS@Standard.
Sun Fire X4200 (6 chip, 12 core)  - 720.56 SPECjAppServer2004 JOPS@Standard.
SPEC, SPECjAppServer reg tm of Standard Performance Evaluation Corporation. All results from www.spec.org as of July 10,2007

 


1 comment:

Stefan Kaltenbrunner said...

Interesting stuff - is there btw a reason why your blog is not aggregated on www.planetpostgresql.org ?