[Discuss] systemd sucks and containers suck and language-specific package repos suck

[markw at mohawksoft.com (markw at mohawksoft.com)]

> On 9/15/26 9:20 AM, markw at mohawksoft.com wrote:
>> I can't think of any large production environment today that is not
>> based
>> on VMs.
>
> Circa 2009 I was working for a company that had a couple of pretty
> crucial production Windows servers, running out of RAM because of bad
> database design, in a colo, and getting crispy such that pushing new RAM
> into one of the machines broke it.
>
> My plan was to quickly put a pair of Supermicro servers there, migrate
> the current software into VMs so they could be given more RAM, moved
> between the two as necessary, be backed up, etc.
>
> When I described the plan to the legacy developer closest to the
> project, he had no idea what I was talking about, but he could tell that
> I was rocking the boat.
>
> I never got the chance to do it. I think that was one of the many times
> I was on the dumped end of a layoff.

At Google, nothing forward facing is real hardware any more. (At least not
that I know of, there may be some weird legacy things that only a few
people know about)

One of the things I should articulate is how amazingly easy it is to
administer and expand this kind of design.

You have between 2 and 2,000,000 "toasters." Any of them can be anything
without any specialized configuration. One of them crashes? The VM is
migrated to another server without missing a beat.

start a postgresql with WAL logging VM, give it some redundant storage,
presto!!! you have a hot-spare postgresql server. Web sites, cluster
compute, the list is endless.  The vanilla cluster with specialized VMs is
the only way to go.