|
Hi,
I'm building a high-availability, high-traffic network and plan to use two
LVS servers as the front end. I havn't had a chance to download the code
and set it up (since I don't have the machines or the time right now) but
was hoping to borrow from the experience of some of the users on this
list.
How scalable is LVS? How many hits can it take and how much does it
reduce throughput? I have the feeling that this is documented somewhere
else, but I didn't find anything that seemed to answer my question
functionally. I actually looked at some of the documentation on the LVS
page regarding performance but it seemed outdated, and in any case
compared various settings of LVS -- I just need some raw numbers versus
hardware (ie 50/hits a second with no noticible lag on this or that
hardware). I would like to hear the opinions of anyone on the list who
has tried using LVS in a high-volume environment.
Also I might need some session awareness -- http, mail, and DB. Does LVS
support some sort of client/server persistance (i.e. load balance but a
given client stays on a given server). One more question (sorry this is
getting so dense): can an LVS redirect an internal request if it protects
multiple networks. To elucidate: if I have a web-server cluster and a
database cluster behind an LVS on seperate subnets can I cause the
web-servers to go through the main LVS to talk to the databases (so they
load balance) or do I need a second LVS between the webservers and the
DBs?
Much thanks in advance,
Sheer El-Showk
|
