[lfs-dev] dbus, systemd, polkit, and consolekit

Bruce Dubbs bruce.dubbs at gmail.com
Sat May 10 09:03:08 PDT 2014

Aleksandar Kuktin wrote:
>> On Sat, 10 May 2014 10:14:14 -0300
>> Fernando de Oliveira <famobr at yahoo.com.br> wrote:
>> [...] who is going to build a server prefers to chose the init
>> system from the beginning and will not be wanting to reboot in
>> different inits. Neither will care about boot/shutdown times, I
>> suspect.
> Strongly disagree. I think one of the reasons systemd is being pushed
> hard by "systems companies" (companies that have a lot to do with
> servers and such) is because of the interplay between servers,
> virtualization, and server instantiation.
> Basically, I believe that, apart from virtualization, hosting companies
> also use server instantiation to promise even more based on even less
> actuall hardware. In the beggining, if you rented a server, you would
> get actuall hardware allocated to you. Then someone invented
> virtualization and then hosting companies would buy less harware and
> run more "servers" on that less hardware, potentially promising
> customers their servers large performances,

Dejavu all over again.  We used to call them mainframes.  :)

but actually being unable
> to deliver everything they promised to everyone they promised it to.
> Now, I think, they are trying to push this even further by litterally
> shuting down virtual servers when they are on a light load. So, I think
> they are trying to make a system where they have few really high power
> CPUs that have many (read: too many) virtual servers allocated to them,
> and they first start one batch of servers, shut them down, start the
> next batch, run them through, shut them down and then go to the first
> batch. All the while promising even more performance to even more people
> while being able to deliver less than they ever were (relative to what
> they promised).

That doesn't make any sense.  I can see copying memory to disk in a form 
of swap or hibernation, but not a shutdown/reboot.

   -- Bruce

More information about the lfs-dev mailing list