|
On Thu, Nov 21, 2002 at 01:42:37PM +0100, Mustapha Hadim wrote:
> Hello LVS list,
>
> I have installed and configured a Piranha cluster (which is an
> LVS subcase) of three real servers and a director, under
> Redhat Linux 7.3.
>
> For the clients, I am using a wireless network. I have a laptop
> PC connected to the Piranha network via a wireless network.
> I have a VMware virtual machine installed within the laptop PC
> client. Thus, I have two clients. I am using http and ftp requests.
>
> That's all fine. What I would like to do, is in fact the following :
>
> I would like to change the scheduling algorithm (until now I am
> using rr scheduling) used in Piranha to take into account the
> network state. Indeed, we would like to implement a new
> scheduling algorithm that will include a monitoring tool. This
> later will give informations about the state of the network, and
> then decisions about the scheduling process will be done.
>
> My question is : where exactly I have to make changes; i.e.,
> in which files ?. Namely, I have to see the source code of
> Piranha scheduling algorithms, and I have to see where to
> include the monitoring tool.
>
> I did't find any documentation about the internal design
> of Piranha.
You should download the source to LVS from linuxvirtualserver.org and
start from there.
Each of the schedulers is implemented as a separate kernel module.
Thus, to implement your new scheduler you will need to implement a new
kernel module. I would suggest copying one of the existing schedulers,
rr (ip_vs_rr.c) might not be such a bad choice, and modifying that. This
should allow you to understand how things work and give you a foundation
to build your scheduler.
--
Horms
|
