|
I suppose a lot of it is the thought that counts. :)
That is -- it'd be nice to be able to add any number of LVSs, even though
most people would stick to two.
And it's not so much that we're worrying about LVSs going down, we're
worrying about the ability to take LVSs down.
One great thing I love about my real server cluster and my redundant LVSs is
that I can take any machine down at any time... do anything I want to it...
and bring it back up (providing other machines haven't failed in the
interim) and no one will know except myself. I like that.
----- Original Message -----
From: "Wayne" <wayne@xxxxxxxxxxxxxxx>
To: "Ted Pavlic" <tpavlic@xxxxxxxxxxx>; "Joseph Mack" <mack@xxxxxxxxxxx>;
"Ratz" <ratz@xxxxxx>
Cc: "Wensong Zhang" <wensong@xxxxxxxxxxxx>;
<lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Sent: Monday, August 07, 2000 6:46 PM
Subject: Re: ideas about kernel masq table syncing ...
> At 06:15 PM 8/7/00 -0400, Ted Pavlic wrote:
> >> Sorry for throwing my 2 cents out, but I think LVS box could never
> >> be the bottle neck for 99.999% situations. If we are thinking
balancing
> >> on the wide area networks, that would be totally different. Another
> >> thought, if the concern was the LVS being the bottle neck, what
> >> about implement something similar to Radware that has both
> >> primary and 2ndary taking traffic at the same time?
> >
> >I don't think Joe was talking about the LVS being a bottleneck. I think
he
> >was talking about having plenty of LVS for high-availability.
>
> Sorry for mis-read Joe's message. Having high quality parts
> in the LVS box can help improving the high availability. Adding
> number of boxes may or may not achieve the high-availability.
> Considering the electronics parts failure rate is a constant,
> the larger number of parts, the higher the failure rate. That is
> why the 20 years old Toyota (with almost no electronics control)
> has much lower failure rate than today's multi-computer controlled
> vehicle (any brand).
>
>
>
>
>
>
>
|
