GF + MySQL + Nehalem = 2925.18 JOPS

ALT DESCR

We have published a new SPECjAppServer 2004 result: 2925.18 JOPS@Standard. Total list price for the configuration, software and hardware together, is $78,834.00, yielding $26.95 per $/JOPS (or, if you are brave enough to go without support, $13.29).

The setup includes GlassFish v2.1 and MySQL 5.1.30 and Sun Fire X2270 and Sun Fire X4170 (both Nehalem-based). Full details in the entries by Madhu, Tom, Kevin and Jennifer.

This is an outstanding result. As Tom describes it:

To give some idea of the actual performance , this configuration of 2 x 8 Core servers and open source , supports a virtual concurrent user load of more than 22,750 web application virtual users and uses > 800 connections to the MySQL 5.1.30 via the Glassfish connection pools!

And this is without the substantial additional improvements from MySQL 5.4, which was not yet available! Stay tuned for further improvements.

The required Disclosure Statement:
SPEC and SPECjAppServer are registered trademarks of Standard Performance Evaluation Corporation. Results from www.spec.org as of 04/20/2009.
Sun GlassFish Enterprise Server v2.1 on Sun Fire X2270 with MySQL 5.1 on OpenSolaris 2008.11. Application Server: 1 x X2270 8 x cores (2 Chips) and Database Server: 1 x X4170 8 x Cores (2 Chips) 2925.18 SPECjAppServer2004 JOPS@Standard

Comments:

Great, I'm happy to see new numbers. I find the JOPS numbers to be difficult to compare with other vendors though because everyone uses vastly different hardware and numbers of nodes. For example:

http://www.spec.org/jAppServer2004/results/res2007q3/

http://www.spec.org/osg/jAppServer2004/results/res2009q2/

WebLogic was 26655 JOPS while GlassFish was 2925 JOPS, but WebLogic was on a 16 server nodes with 64 instances. GlassFish was on on one server with 4 instances. This kind of testing makes JOPS seem a bit useless to me since I can't compare vendors.

Posted by Ryan de Laplante on April 24, 2009 at 06:06 PM PDT #

It's how the benchmark is designed. It is mostly a hardware benchmark and it is not good particularly good at measuring apples to apples, although it is an indication of how well you can scale the thing (and also how stable).

To get some apples-to-apples comparison we started counting $/JOPS. You can compare HW-$/JOPS or TOTAL-$/JOPS and get some sense of value.

- eduard/o

Posted by Eduardo Pelegri-Llopart on April 25, 2009 at 01:11 AM PDT #

Post a Comment:
Comments are closed for this entry.