Docker ascendancy's ignites a flak-in-the-box cloud arms race

Containerisation has taken the data centre by storm. Led by Docker, a start-up that's on a mission to make development and deployment as simple as it should be, Linux containers are fast changing the way developers work and devops teams deploy. Containerisation is such a powerful idea that it's only slightly hyperbolic to …

[SiggyMax]

Why is it that ...

we again and again must hear that the traditional OS is dead? This article states: "To be sure it's still a ways off, but containerisation is likely to completely replace traditional operating systems – whether Linux, Windows, Solaris or FreeBSD – on servers. Instead, servers will consist of simple, single-user installs of hypervisors optimised for the specific hardware".

But the point with Docker is the direct opposite - it replaces the hypervisor, that does a lot of heavy work emulating a full hardware platform, with a single instance of a host OS (Linux in Docker's case) shared between multiple containers. Something that allows quite a lot more applications to be run on top of the iron at the bottom, which (together with the ease of management/installtation) seems to be the reason for Docker's success.

Each time I hear that "the OS is dead, hail the Next Big Thing" I eagerly search for the details about how The Next Big Thing manages to replace memory management, file systems, networking stacks, user and security management and so on and so forth, all the stuff applications rely on, with absolutely nothing. And so far the answer has been: The Next Big Thing doesn't. It just wraps the Good Ole' OS in a thin veil of hype and buzzwords.

[Anonymous Coward]

You should have considered extending the analogy a step farther. One of the ship's Purser's roles is to properly place the containers, guided by weight, so that the moment arms are all in the right placement to enhance the stability of the ship rather than enhance the ship's ability to capsize. Capsizing is very bad. Docker, and containers in general, do have flaws and the extension here is the load demands must be balanced properly. I'd even extend it to having proper hardware support (requirements) should the container ask for it (and fail/degrade properly). There are loads, automobiles as just one case, where a different ship (Whale) may be required.

I love the concept, it's something I've been advocating since the late '80's but like AJAX, it had to simmer a long while before a use was found for it. [XMLHTTPServerRequest() and XMLHTTPRequest() were available back in '95.] This is another attempt not only to create write-once software but a chance to standardize the components (containers). Done right, it could be very, very nice. I don't hold much hope though: I'm expecting it to be another great white hope of a whale. Very Moby Dick.

[nematoad]

?

"Rocket offers some real advantage over Docker...arguably, more in line with the popular Unix philosophy of small parts loosely joined."

Maybe not if it is using systemd. The whole rationale behind systemd is to integrate everything and make it so that everything has one point of contact.

A good discussion of the use of systemd with servers can be found:

Here

[Anonymous Coward]

How are the normal important things dealt with, application users, system users, keys, passwords, security patching, firewall configuration, IPS, IDS, routing, etc managed? As in my lengthy experience with developers the normal fix (is the equivalent of) set it to 777 and tada all the problems go away.

A single user to do everything, firewalls turned off, no selinux, no tripwire/ossec, no patching, a shitty password and the same key in prod as in dev everything on the same network tier (or more often the same box because the actual development environments are "too hard" to use...)

I don't begrudge them (bar the many many hours of my life spent trying to figure out why some pos thing doesn't work and it's coz the dev had done something ridiculous) they want to get things out the door and have deadlines and working with people that actually understand how systems work can be a royal pain in the arse (they're normally busy trying to beat the application in production back into submission.) Sure some devs are awesome with systems, others barely know how the internet works.

So yeah - how is all that kind of stuff dealt with by docker - in the real world as opposed to the land of start ups and new modern applications? CenturyLink keep trying to run their docker lard all over us and anything they try and sell us is invariably shitty half backed penis.