|
On Sunday 01 July 2007 11:58, Joseph Mack NA3T wrote:
> On Fri, 29 Jun 2007, Rio wrote:
>
> > also please check out linux-vserver
> >
> > http://linux-vserver.org/Welcome_to_Linux-VServer.org
> >
> > we have been running it for a year now with absolutely no hiccups
whatsoever
> > and no excessive loading! we have 84 virtual servers on 1 machine and 40
on
> > another machine and i am configuring a third host as i write this. my
> > estimates are that the 2 existing machines could easily handle 100 virtual
> > servers each. the virtuals can be mostly any linux distro mix though the
most
> > popular are gentoo, debian, redhat, centos and ubuntu.
>
> Trying to find out why people use virtual servers for
> realservers
>
> One person here (forget who) said that it was cheaper to
> have one big server than to have the same server capacity in
> single machines.
>
> The next question then is why don't you have the big server
> which is currently split into 84 realservers as just one big
> realserver? If you need failover (do virtual servers fail?)
> why not just 5 machines (enough that you'll only loose 20%
> on failure)?
if i am understanding you to mean you want me to split it into 84 complete
virtual machines that advertise their own virtual hardware? too much overhead
and wasted resources. we do have several instances where virtualized hardware
is required. we have one host running esx server for just that need. but by
far the most common needs are easily met using linux-vserver.
if i understand you to mean combine the 84 servers into one server running all
the services, that cannot be. some of those are rented colo space and i will
not allow a customer access to our hosts. also, if someone does something
stupid that takes their virtual server down, no loss to us, we simply keep
the mistake from running and restart the server. our services follow certain
classifications and we divide those into virtuals for ease as well as of
course customers having their 'own'. if it was one big server they would
take the entire system down. no way. i have become a firm believer in context
virtuals. far less overhead than virtual hardware (who needs 84 kernels
running all doing basically the same thing?).
maybe i completely misunderstand the question in which case i apologize for
being brain-dead :)
virtuals never 'fail' unless someone with access to a particular virtual does
something stupid. however occasionally it may be necessary to take a virtual
down for a bit. failover is nice in that instance as well as an aside
function of failover can be round-robin or other methods of balancing loads.
>
> Thanks Joe
>
>
> --
> Joseph Mack NA3T EME(B,D), FM05lw North Carolina
> jmack (at) wm7d (dot) net - azimuthal equidistant map
> generator at http://www.wm7d.net/azproj.shtml
> Homepage http://www.austintek.com/ It's GNU/Linux!
>
> _______________________________________________
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
>
--
Rio
|
