> On 30/12/2015, at 12:11 AM, a b <[email protected]> wrote:
> And what do people do with it? Run fully virtualized systems with
> overcommit set to 63% of memory and 263% of central processing
> units' capacity. It does not run great, but it runs "good
> enough" for VMware to dominate the segment.
More to the point it runs well enough for their clients. If you're an SME with
half a dozen Windows centric admins on staff, "we can still run AD" beats any
theoretical advantage into a smoking hole. Furthermore any solution must be GUI
managed. So they could do it with SmartOS and KVM's but it would take forever
and a day, would annoy the staff intensely and be a complete bitch to hire for.
Ten grand in VMWare licenses is a much better option.
Nobody ever got fired for choosing VMWare.
> SmartOS does not work for ${INSURANCE_COMPANY} or ${BANK}
Well, they have Solaris because nobody ever got fired for choosing that either.
> SmartOS was
> absolute taboo because the system engineering department was com-
> pletely unfamiliar with it, and because it was perceived as pol-
> itically incorrect to even so much as mention anything that was
> not VMware and Linux.
Well, yes, but I can see their point. You can't buy SmartOS support and this
alone is enough to stop it being a contender in 99% of cases. Besides, you know
that VMWare and RedHat have been tested together and that there's probably a
million instances running at any one time. Most people really don't like
dtrace'ing problems and would rather their stuff just worked.
> People around me, all with Bachelors or Masters in computer sci-
> ence could not even program and did not even hear of things such
> as the AWK programming language
I only discovered awk a couple of weeks back. It is kinda handy but if one is
an Objective-C developer it's not something that would appear on the radar.
>> It also
>> assumes that we're trying to build a platform on which every legacy
>> application that has ever existed can be run in some fault tolerant
>> configuration without any work.
>
> And why not? It is a problem in the OS domain which has yet to be
> solved.
But it's (a) really, really hard and (b) doesn't need to be solved because
we're doing that with containers now. Or, will be soon.
>
> So if I understand you correctly, it does not bother you if Smar-
> tOS does not become a mainstream adopted product?
No, not at all. The only thing that concerns me is if one day Joyent decide to
stop throwing resource into SmartOS development and ... well, we are where we
are. Particularly if there are niggly, hard to reproduce problems with running
docker containers then I might be really kinda stuffed. If I were working for
someone when this happened I might not be fired, but I'd almost certainly lose
the right to make decisions.
> I do not; good old DNS and SRV records will do quite nicely for
> service discovery.
They get cached, unfortunately. You can set a low timeout but an application
will do the lookup once then assume it's there forever. You also can't 'push'
changes. That being said I am a huge fan of old school technology because
'tested to death' and 'really simple' appeal greatly.
-Dave
-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription:
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com
