Re: IPVS scheduler algorithms (was: [PATCH] ipvs: Kconfig cleanup)

On Sun, Jul 11, 2010 at 07:15:21PM -0500, Ismael Luque Valencia wrote:

Hi my name is Ismael, I am doing a paper about lvs algorithms but I dont
undernstand two of them Locality-Based Least-Connection and Locality-Based
Least-Connection with Replication
Someone, who can explain me that algorithms please
If it is in Spanish will better because my English is not good

Hi,

Firstly, please don't reply to emails unless you are actually
replying to the topic at hand. Instead just compose a fresh
message to the people/lists that you want to address.
This helps keep threads in mail-readers that support them sane.

In any case this question would be better sent to the lvs-users list.

But to your question.

These schedulers are described briefly in the ipvsadm(8) man page
and in pseudo code in the source files (ip_vs_lblc.c and ip_vs_lblcr.c)
in the Linux kernel tree. And there is some discussion in the howto
http://www.austintek.com/LVS/LVS-HOWTO/HOWTO/LVS-HOWTO.ipvsadm.html#DH

The way that I think of these schedulers is as enhanced versions of wlc
designed for use with transparent proxies. That is situations where
there will be a lot of destination addresses.

lblc works by keeping a cache that associates destination addresses with
a real server. This allows accesses, potentially from different end-users,
to be sent to the same real-server. As this is designed to be used
with proxies, this means the request will be sent to a proxy that may
have already retrieved the result.

lblcr is similar, but instead of one real-server per destination, it
allows for multiple real-servers per destination.