[Info-vax] Data Center Operations: DC/OS, Apache Mesos

[Paul Sture]

On 2016-06-03, Stephen Hoffman <seaohveh at hoffmanlabs.invalid> wrote:
>
> There have been suggestions here in the comp.os.vms newsgroup that 
> OpenVMS head toward a distributed data center operating system, and 
> there have been a few discussions both here in comp.os.vms and at the 
> 2015 OpenVMS boot camp around adding support for containers on OpenVMS 
> (both as containers and in the guise of app stacking).

A very timely post, as I was building myself a test environment to
go through the SaltStack tutorials when it arrived :-)  [1]

> If that sounds interesting, have a look at Mesoshere's DC/OS:  
> https://dcos.io/docs/1.7/overview/what-is-dcos/
>
> As dealing with all that is part of what Apache Mesos — the 
> underpinnings of DC/OS — provides.  http://mesos.apache.org

Thanks for the links.

> Through Mesos, DC/OS allows scheduling on local boxes through various 
> means, as well as on AWS and Azure services.
>
> FWIW, the "When you’re operating a datacenter, there are a set of 
> common operations that you do because you have to, not because you want 
> to. In a single computer environment, your operating system 
> automatically takes care of these things. When was the last time you 
> had to manually tell your laptop which processor core to run your 
> application on?  For those of you who responded with anything other 
> than “never”, there are some amazing computers at the Computer History 
> Museum that you might want to check out." line was good for a chuckle.

Just last year when I was mucking around with scientific stuff.  Before
that, 1997 when pinning a CPU hog of a 16-bit app to a second processor
so that workstation tasks could continue as normal on the primary
processor (OK, a tower system rather than laptop...).

> No, I haven't seen a port of Mesos to OpenVMS, and I'd tend to expect 
> the down-standard C++ support on OpenVMS probably lacks a few features 
> necessary for the port.
>
> But maybe this gives some of you some idea of where parts of the 
> high-end and data center computing markets are already headed.

Yep.

[1] <https://saltstack.com/community/>

As an aside, the SaltStack tutorials suggest using Vagrant.  It's more
than a year since I tried Vagrant but I found it bandwidth hungry to the
point where that aspect, for me, outweighed the advertised advantages of
the product.

Slight (Vagrant inspired, but Vagrant is far from the only guilty party
here) rant follows.

I don't care how fast your internet connection is, there's still no
point in repeatedly downloading the same stuff again and again.  If
you've got the bandwidth to saturate the spinning rust disks I have
here, you've probably got much faster I/O systems as well and you're in
the same boat as I am.


-- 
There are two hard things in computer science, and they are cache invalidation,
naming, and off-by-one errors.