Hi,
I've read and heared lot about the low mem consumption of OpenVZ VE's. Stories about running 200 VE's (with only apache, and usual stuff) on 2GB of RAM.
When I have a look at mine machine I can tell you this is impossible!
If I execute the top command within the VE's each VE consumes at least 140MB of RAM.
One of the VE's is for example the debian-joomla template from proxmox (apache, php5, mysql, 512MB Ram assigned). This VE consumes (at idle) 140MB. When under little load (testing a simple php page on apache with ~50 concurrent users and ~10% cpu utilization) the mem consumption goes a little above 200MB.
If I assign 200MB of RAM to each comparable VE this means I can only have 10-12 VE's on a machine with 2GB RAM (host's mem consumption not taken into account).
200 vs. 12!
How can one have 200 VE's running on 2GB RAM??
Do I have to tune something to achieve lower mem consumptions?
Has anyone else running the debian-joomla template? What is your (idle) memory rate?
I've done some (simple) tests with different OpenVZ and KVM templates measuring the cpu and mem rates using simple Java app tests on Tomcat.
KVM VE's appeared to consume between 2-4 times more CPU then OpenVZ templates. Mem consumption was about the same BUT with each n'th test the mem usage kept increasing on KVM templates while OpenVZ templates stayed about the same after 3rd run. I can send my spreadsheet with the test data to anyone who is interested.
I'm using: pve-manager/1.0/3463 on a Quad Core Xeon with 12GB RAM.
Note: I don't have to deploy such high number of VE's running on my machine. It's just curiosity.
I've read and heared lot about the low mem consumption of OpenVZ VE's. Stories about running 200 VE's (with only apache, and usual stuff) on 2GB of RAM.
When I have a look at mine machine I can tell you this is impossible!
If I execute the top command within the VE's each VE consumes at least 140MB of RAM.
One of the VE's is for example the debian-joomla template from proxmox (apache, php5, mysql, 512MB Ram assigned). This VE consumes (at idle) 140MB. When under little load (testing a simple php page on apache with ~50 concurrent users and ~10% cpu utilization) the mem consumption goes a little above 200MB.
If I assign 200MB of RAM to each comparable VE this means I can only have 10-12 VE's on a machine with 2GB RAM (host's mem consumption not taken into account).
200 vs. 12!
How can one have 200 VE's running on 2GB RAM??
Do I have to tune something to achieve lower mem consumptions?
Has anyone else running the debian-joomla template? What is your (idle) memory rate?
I've done some (simple) tests with different OpenVZ and KVM templates measuring the cpu and mem rates using simple Java app tests on Tomcat.
KVM VE's appeared to consume between 2-4 times more CPU then OpenVZ templates. Mem consumption was about the same BUT with each n'th test the mem usage kept increasing on KVM templates while OpenVZ templates stayed about the same after 3rd run. I can send my spreadsheet with the test data to anyone who is interested.
I'm using: pve-manager/1.0/3463 on a Quad Core Xeon with 12GB RAM.
Note: I don't have to deploy such high number of VE's running on my machine. It's just curiosity.