Docker, Part 2: Whoa! Spontaneous industry standard! How did they do THAT?

Docker is slowly taking over the world. From its humble origins, which we explored on Friday, as an internal project at dotCloud, through to Microsoft's recent announcement that it will support Docker natively in Windows, Docker looks set to become a major component of modern IT infrastructure. Today, Docker is powered by …

[Graham 24]

Change of heart?

Can I ask what's prompted the change of heart from

"I might write it for Docker, once Docker has things like FT, HA and vMotion, but I'm honestly not sure why I'd bother, Docker seems like more work than AWS and doesn't offer a fraction of the flexibility you get when using a proper hypervisor."

(http://www.theregister.co.uk/2014/09/17/pay_for_what_you_provision_cloud/?page=2)

to today's

"Docker looks set to become a major component of modern IT infrastructure."

As far as I can see, Docker still doesn't have FT, HA or vMotion-esque features (unless you consider those to be present one level up, at the parent OS level). Are they coming in some future release?

[Trevor_Pott]

Re: Change of heart?

There are two more installments in this series.

But, if I may, please understand that I am aware that there is more to our industry than "Trevor Pott and people who think like Trevor Pott." I absolutely won't be using Docker until it has things like FT, HA and vMotion...but I am largely a keeper of "legacy" workloads. Traditional applications; not the sort that are optimized for, for example, cloud computing.

Those with money - startups with VC funding, governments and enterprises - absolutely are rewriting extant applications to take advantage of "cloud" technologies. These recodes translate almost directly to being "good for Docker". Also, a huge chunk of all new application development follows that model.

In the old world of the sorts of "legacy" applications that I herd - back when applications were applications, not "apps" - you would have a few components to worry about: file storage, database, the application itself and the client. Eventually "the application itself" and "the client" became more or less one thing as things went to web-based applications. But we still had these three things that absolutely had to be up 100% of the time. Both scale up and scale out were (and remain, for us legacy herders,) A Great Big Bitch Of A Problem.

"Redundancy" comes from VMs, or from NonStop servers. The database isn't allowed to go down. Clusters only work if your database app supports it, and you usually have to convince the vendor (who may not even exist anymore) to recode some chunk of the app/database. If the devs that are left even know how to do that!

Modern "apps" are totally different. They're built from the start to be able to scale out and up. It can collapse down to one core copy of the DB/Files/App or scale out to thousands.

In a modern app, you only need to keep the master copy safe. Everything else can spawn some unlimited number of copies as needed.

In Theory. Truth is, doing so in practice is still A Bitch, but it's not quite A Great Big Bitch.

Then along comes Docker. Docker makes the "scale out" part of the modern apps thing Creepy Meerkat Simples. That's grand. So if you want to run Netflix-class infrastructure, you can basically put your core stuff on a NonStop server, then spawn unlimited Docker instances out on a bunch of cheap metal. AMD SeaMicro, HP Moonshot or Supermicro MicroCloud, anyone?

Voila: a use case for the next generation, albeit not one that I will myself be using any time soon. And - just by the by - it's a use case worth hundreds of billions of dollars. Mind you, so is keeping those legacy workloads running.

Our industry is diverse. And Docker has added to the choices before us. It is only one tool in the toolbox. It is kinda odd and non-standard to us old fogies...

...but I promise you, it will be the #2 Philips Screwdriver of the generation that's just cutting their teeth today.

Edit: let me add that the above should read containerization will be a multi-billion-dollar industry. Whether or not Docker, specifically wins this war is as yet undetermined. That said, containerization's time has come.

[proud2bgrumpy]

Re: Change of heart?

"As far as I can see, Docker still doesn't have FT, HA or vMotion-esque features (unless you consider those to be present one level up, at the parent OS level). Are they coming in some future release?"

- Yes, I think that is exactly where those feature belong (actually with the Host OS) - a VM really shouldn't have the level of access to the Host to initiate such features, so a Container (which is itself one level further abstracted from the HW) shouldn't either. How would a VM or a Container know which Host it should move to? How could it know what environmental configuration settings to change?

[Trevor_Pott]

Re: Change of heart?

...

What?

Where/how did you pull from anything above that the application or guest OS should be the one moving things around? Whaaaaa?

[Phil_Evans]

Under-played, or future reading?

Great series, narrowing the gaps between what I think I know and mea culpa non-scooby.

I wonder if you're likely to talk more about why Microsoft is such a 'Yes Man/thing' in the pre-flight phase? I see the business model of a host of hosters dissolving readily under Docker's arguments since the whole 'host' thing become a bit of an irrelevance. Given that Azure now runs (almost) any flavour of 'app' that the hipsters desire, the need for the shiny expensive Azure stack itself seems redundant in the end. Same for AWS' Windows tie-up?

For Linux-hosters, this is but a consideration of consolidation, surely?

There may still seem a gap between what I think I know and....etc.

[Trevor_Pott]

Re: Under-played, or future reading?

"Watch this space". ;)

If you've questions after the fourth article is out, I'll be glad to fill in the gaps. Cheers!

[Victor 2]

MySQL

You should update yourself from the FUD after Sun's acquisition. MySQL as a product and as community is doing quite alright now.

[M Gale]

almost nobody says "Android Linux."

And that's because once you add on the Dalvik/ART stack and everything else that goes with it, Android is about as much Linux as Java or Mono/.Net is: It isn't.

Does anybody say "TiVo Linux"?

[ratfox]

Re: almost nobody says "Android Linux."

Well, Linux is just the kernel, doncha know…

[Paul]

I'm not sure whether this article had really all that much content.

[John Sanders]

So far two articles

And nothing new that we did not already knew.

That Docker is hyped to the max by people who struggle with system administration (Modern young developers, excitable people who believe in buzzwords like "continuous delivery" as if that was something that leads to quality software.)

That the entire industry likes Docker (Because the industry loves it, period)

That solves some problems, but it is not appropriate for all work-flows (many which can not simply be run inside a container)

And that the company making it is positioning themselves as another "benevolent" dictator amidst all the hype.

So far the only use for docker that I have seen was to deploy a php application, the guy who showed me was overly excited about the fact that he did not have to fiddle with Apache's config, just download an already made container with the application run and profit.

I realize that to some people that could be a big thing, (the people who like clicking next-next-next, I have no clue what's going on behind the scenes.) Maybe I'm old and having to edit Apache config files, installing dependencies, and doing some troubleshooting ts not as scary to me as it is to them, but I get the feeling that it will be me the one who will have to fix the containers when they break and doing next+next+next does not fix the problem.

This reminds me of the early 2000's Microsoft .net tv ads, in the ads .net seemed to be able to grant legacy applications functionality that they did not posses "I did not know this system could connect to this system", The old beards in the company I was working for at the time were puzzled as to how could .net perform the miracle.

Docker is the same, it is another technology on the arsenal, its not going to do anything magic, but to the people for which everything looks like a nail Docker is a hammer.

[Anonymous Coward]

I'm not sure whether this article had really all that much content.

I understood very little of it. It read like an HP whitepaper, at the end of which you still don't know what the product actually does. I realise that for the people who do, the article is probably very meaningful and insightful but...I'm glad I'm retired.

[Cipher]

Docker, like it or not?

Poettering is in on this then? I guess even Android isn't safe from systemdesque thinking...

[Destroy All Monsters]

Re: Docker, like it or not?

It's just like another Jason/Freddy installment.

[Anonymous Coward]

Part 1?

Any chance of a link to part 1 of this?

[Trevor_Pott]

Re: Part 1?

Part 1

Part 2

Part 3

Part 4