[#PETALSESBCONT-72] Check performance consistency between a small 32 bits server and a powerful 64 bits server - Petals Link JIRA

Details

Type: Task
Status: New
Priority: Blocker
Resolution: Unresolved
Affects Version/s: 3.0.5
Fix Version/s: None
Component/s: None
Security Level: Public

Description:
Hide
The Petals topology used is the default one (standalone mode, Petals ESB 3.0.5, JDK 6)
The BC SOAP (4.0.2) and SE POJO (2.2.2) are installed (the XSL transformation and backup are packaged into a JAR to prevent side effects)
The servers used are:

32 bits: 1.7 GB memory/1 EC2 Compute Unit (1 virtual core with 1 EC2 Compute Unit)/160 GB instance storage

64bit: 68.4 GB memory/26 EC2 Compute Units (8 virtual cores with 3.25 EC2 Compute Units each)/1690 GB of instance storage
Note 1: An EC2 Compute Unit is same as a 2007 Xeon processor
Note 2: Operating system: Linux Debian, the last version available in 2010, September/October.

The 32bits server is more powerful than the 64bits:

32 bits: 150 msg/s

64 bits: 20msg/s

(Thanks to Nabil Sid, Logica).
Show
The Petals topology used is the default one (standalone mode, Petals ESB 3.0.5, JDK 6) The BC SOAP (4.0.2) and SE POJO (2.2.2) are installed (the XSL transformation and backup are packaged into a JAR to prevent side effects) The servers used are:

32 bits: 1.7 GB memory/1 EC2 Compute Unit (1 virtual core with 1 EC2 Compute Unit)/160 GB instance storage

64bit: 68.4 GB memory/26 EC2 Compute Units (8 virtual cores with 3.25 EC2 Compute Units each)/1690 GB of instance storage Note 1: An EC2 Compute Unit is same as a 2007 Xeon processor Note 2: Operating system: Linux Debian, the last version available in 2010, September/October.

The 32bits server is more powerful than the 64bits:

32 bits: 150 msg/s

64 bits: 20msg/s

(Thanks to Nabil Sid, Logica).

Environment:

See description

Download all attachments as a ZIP file

Attachments

32bitsVs64bitsProject.zip

(47 kB)

Christophe DENEUX

Fri, 15 Oct 2010 - 12:29:45 +0200

Activity

Ascending order - Click to sort in descending order

Hide

Permalink

Christophe DENEUX added a comment - Fri, 15 Oct 2010 - 12:29:45 +0200 - edited

Sas and project sources

Show

Christophe DENEUX added a comment - Fri, 15 Oct 2010 - 12:29:45 +0200 - edited Sas and project sources

Hide

Permalink

Sébastien André added a comment - Thu, 28 Oct 2010 - 11:26:54 +0200 - edited

La difference de performance n'est pas forcement due a la difference d'architecture des CPUs.

Si "virtual core" correspond bien a l'affectation d'un CPU pour une machine virtuelle dans une baie, c'est probablement le nombre different de CPUs dans chaque test qui change les performances.

Avec 8 CPUs le temps "perdu" pour paralléliser est important et est tres dependant du test.
Si le test est surtout du transfert de message, ca ne se parallelise pas bien

Ce serait interessant d'avoir le resultat du test avec le meme nombre de virtual cores pour chaque architecture de CPU.

Show

Sébastien André added a comment - Thu, 28 Oct 2010 - 11:26:54 +0200 - edited La difference de performance n'est pas forcement due a la difference d'architecture des CPUs. Si "virtual core" correspond bien a l'affectation d'un CPU pour une machine virtuelle dans une baie, c'est probablement le nombre different de CPUs dans chaque test qui change les performances. Avec 8 CPUs le temps "perdu" pour paralléliser est important et est tres dependant du test. Si le test est surtout du transfert de message, ca ne se parallelise pas bien Ce serait interessant d'avoir le resultat du test avec le meme nombre de virtual cores pour chaque architecture de CPU.

Petals ESB Container

Check performance consistency between a small 32 bits server and a powerful 64 bits server

Details

Attachments

Activity

People

Dates