SPECjbb2005: Multiple Instance Results

SPECjbb2005 can be run in a multiple-instance mode, where n JVMs run in parallel, with the benchmark load distributed between the separate JVM processes. Take a look at the first ever Multiple-Instance result here: http://www.spec.org/jbb2005/results/res2005q2/jbb2005-20050613-00007.html The first question you might ask is: What's bops and bops/jvm? Here's the scores listed: bops = 75,862 bops / JVM = 18,966 bops is the total benchmark score for the system. It is the sum of the work performed by each of the individual JVMs. bops / JVM is the average work performed by each individual JVM. Now, how does this compare to a single instance result? Take a look at our Sun Fire V40z result: http://www.spec.org/jbb2005/results/res2005q2/jbb2005-20050613-00003.html Here's the scores listed: bops = 37,034 bops / JVM = 37,034 bops and bops/JVM are identical. Makes sense since this is a single instance result and therefore there is only 1 JVM. So how do you compare single and multiple instance results? First a bit of background. The single instance results highlight hardware, OS, and JVM performance and scalability. The multiple instance results highlight hardware, OS, JVM performance and scalability, and JVM interoperability. Java is deployed in the enterprise in many different ways and a common production environment is a large SMP server running many (10s to 100s) JVMs. The multiple instance configuration allow JVM and hardware vendors to demonstrate this common deployment models in a public competitive benchmark, especially useful on large SMP servers. This was a great idea. So again, how do you compare single and multiple results? Use the bops score alone. A comparison between these two configurations ends up being a limited comparison of system scalability and performance. Limited in that hardware and OS scalability is tested, but JVM scalability has been limited to include the effect of many JVMs working together. So here's a comparsion on the Sun V40z 4-way result and the Sun V890 8-way results: 4-way V40z: 37034 bops 8-way V890: 75862 bops
Comments:

So I guess this would indicate that a 490 is as fast as a v40z? (Half of the 890?)

Posted by KG on June 26, 2005 at 08:02 AM EDT #

Yes, based on the V890 results and this benchmark. As a perf. guy I wouldn't say for sure without testing..

Posted by Dagastine on June 26, 2005 at 01:03 PM EDT #

Post a Comment:
  • HTML Syntax: NOT allowed
About

dagastine

Search

Archives
« July 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
  
       
Today