|
|
Re: load balancing and migration between hardware nodes. [message #4801 is a reply to message #4772] |
Wed, 26 July 2006 16:14 ![Go to previous message Go to previous message](/theme/ovz3/images/up.png) ![Go to next message Go to next message](/theme/ovz3/images/down.png) |
edbch
Messages: 5 Registered: July 2006
|
Junior Member |
|
|
First, thanks a lot for answering so readily. Second, sorry my bad English.
Yea, I want to migrate a VE from one hardware node to another.
But it would like to make this in automatica way. I would like to use some load to balancer to distribute the VEs for mine we in my farm, thus to distribute better the load for them.
It assumes this scene: I have 30 VEs in three machines. Of these the 30 three first ones are runing applications that they demand much cpu or memory. I would like, then, to use some load to balancer so that these three first ones were distributed in the three machines, without I to have to make manually this.
Thanks a lot any help.
Eduardo
[Updated on: Wed, 26 July 2006 16:15] Report message to a moderator
|
|
|
Re: load balancing and migration between hardware nodes. [message #4810 is a reply to message #4801] |
Wed, 26 July 2006 23:48 ![Go to previous message Go to previous message](/theme/ovz3/images/up.png) ![Go to next message Go to next message](/theme/ovz3/images/down.png) |
|
There are a lot of factors to be considered to decide what, when and how to migrate, that's why an automatic solution would be too complex and therefore looks like a dream. You tell that CPU is a concern -- but other than CPU, there are other resources like memory, disk space, I/O load, network bandwidth etc. Note that the migration itself is also consuming some resources -- and that needs to be taken into account as well (or it might end up migrating VEs back and forth and keep the system busy with just that).
The best thing that could be done in that regard is to write some code which will analyze the situation on all the nodes (using vzcpucheck, vzmemcheck, probably iptables statistics, load average in all VEs, and tons of other statistics that could be gathered) and suggest some improvements to the human which can than decide what to do.
These are just some basic initial thoughts. If you are a programmer, you might want to try implementing some of that (or hire somebody to do that if you are not).
Kir Kolyshkin
|
|
|
|
Re: load balancing and migration between hardware nodes. [message #4840 is a reply to message #4810] |
Thu, 27 July 2006 17:38 ![Go to previous message Go to previous message](/theme/ovz3/images/up.png) ![Go to next message Go to next message](/theme/ovz3/images/down.png) |
edbch
Messages: 5 Registered: July 2006
|
Junior Member |
|
|
First, thanks for answering so readily.
Sorry if I did not point out oproblema with details. I really have some situations whith a problem specify and distinct, but when I imagined the solution for my problem I thought about as to decide this in generica way. In fact thinking about the solution I presumed that as already had a tool to migrar itself thought that perhaps already an prototype of some species of loadbalancer existed, or one frontend for one that exist, with features of configuration for diverse scenes. Without a doubt something as this must take many factors as you yourselves it said, some minimum things for example minimum time to remigrar a VE. As it said I presumed that the development team could already have some and that I, I eat programmer, could contribute. I find that I presumed excessively.
Well, if this does not exist, finds I could think to start something to decide my problem of start, even because I find it is not so rare. My next step would be to ask for programs (beyond that you cited) to get necessary information so that “my” loadbalancer it decides correctly on when migrar a VE.
Thanks again for your time.
|
|
|
Re: load balancing and migration between hardware nodes. [message #4841 is a reply to message #4836] |
Thu, 27 July 2006 17:47 ![Go to previous message Go to previous message](/theme/ovz3/images/up.png) |
edbch
Messages: 5 Registered: July 2006
|
Junior Member |
|
|
Tanks for answering. As I said for the Kir, I really did not specify my problem because he was thinking about a solution of generic form, exactly knowing that they are many variables. As I said I imagined that perhaps already an prototype existed for that it allowed to make configurations for specific situations. As seems that this does not exist I go to try to decide my problem with what I have available.
Tanks again for its time.
|
|
|