Private cloud and OpenStack: You ain't heard nothin' yet
And "watch" is the key word here, since M was Lang's first sound film, and he was still tentatively exploring how to best make use of the still brand-new dimension of audio. In fact, there are long stretches in M during which there is no dialogue, no ambient sound, no soundtrack, just silence. That said, he experimented with a number of techniques that were far beyond what was being done in the early "talkies", but he still wasn't yet comfortable enough with the medium to completely rely on sound; instead he relied on what he knew well and made what is a visually gripping film.
But what does this have to do with OpenStack? More than you think.
Imagine a classic datacenter, with racks of servers running legacy applications; or a first-generation virtualized datacenter, where those same apps run on highly-resourced, carefully hand-configured virtual machines. That's "Silent Film."
Then imagine someone comes along and says: "Well, now you can toss apps into containers and start them in milliseconds and scale them out and load-balance them and launch them from smart, simple, dynamic templates and scripts that can then be automated further to scale entire clusters of apps across multiple servers in response to traffic demands, all without human intervention." But then imagine they also say: "The catch, though, is that you have to write your apps in a new way for this all to work. And your legacy apps, well, they'll still work, but they won't benefit directly from all this new magic." That's "Sound in film."
There's nothing inherently wrong with legacy infrastructure. Like silent films, it does what it's supposed to do -- often with enormous power. Add cloud technology to the mix, and it can do so much more. Consider what Star Wars, or The Avengers, or even a drama like Birdman would be like without sound. They simply wouldn't exist.
That's not to say, though, that the transition was easy. Sound was as much a disruptive force to films as cloud technology is to IT today, and it takes time to adjust.
When Boris Renski gave a keynote address to the OpenStack Summit in Austin last month, he led off with last year's Gartner study that talked about the problems that affect 95% of private clouds. Only 6% of those problems, he pointed out, were rooted in the technology. The other 94% was all either people or process.
The largest percentage of problems, for example, was "failure to change operational model", at 31%. He described cloud as a stack, with the cloud technology at the bottom, operations in the middle, and the tenants on top.
That middle part is where things have to change, he explained. In a public cloud, that middle portion is DevOps ninjas who run around putting out fires and making things run smoothly for the tenants. In a private cloud, this is usually VMware admins who haven't quite made the switch to this "on demand" mindset.
"If you're embracing OpenStack as just technology," he said, "you will probably fail. Nowadays we try and tackle the people problem and the process problem." When working with customers, the company looks for both an enthusiastic sysadmin to work with and an anchor tenant to work for, then defines success metrics for the anchor tenant. If it's just the admin's project, he explained, it doesn't help anybody (except the admin and his/her resume).
It's kind of like the first talkies. They were essentially filmed plays, talking heads that relied on the novelty of synchronized voices to enthrall the audience. Think of that as the equivalent of forklifting your traditional app onto a cloud-based server. Yes, you're "using cloud", but that's about all you can say for it; you're not doing anything new or innovative; you're not really taking advantage of what the new medium has to offer. After a while the audience gets bored.
So how do you solve this problem? Using private cloud effectively, Renski explained, means turning traditional administrators into DevOps ninjas. In today's world, the only other alternative is to keep doing things the way you always did, and go to public cloud. But using public cloud doesn't solve the people and process problem, it just shortcuts it; the cloud doesn't magically stay up all the time without DevOps ninjas, it's just that the process is hidden.
It seems black and white, he says. Public cloud or private cloud. But there really are a bunch of different options on the spectrum, such as managed private cloud or appliances or any number of options that are just a tiny sliver of the mechanisms that will eventually appear.
In other words, just as sound created a whole new form of entertainment when it came to movies, AWS created a whole new paradigm for computing -- and neither is an instant transition.
Let's go back to our Fritz Lang example. M, as I said, is a visually stunning film, and frankly, most of it could probably have worked as a silent film. There are images that are still with me long after I finished watching. (You'll never look at a balloon the same way, trust me.)
But if traditional infrastructure (and emulating it with public cloud) corresponds to silent films in this analogy, does talking about the visual quality of M that mean I'm advocating sticking with what you know, and what works well?
Absolutely not. While M is a gripping thriller throughout, what really makes it a classic is the part of the film that is probably the reason Peter Lorre became a star: his speech at the end. This is a sequence that could not possibly have been accomplished without sound. Even listening to him in German while you read the subtitles, you can't help but marvel at how Lang took what had, just four years earlier, been a silent medium, and used sound to its best effect to create something that even now, 85 years later, keeps you thinking about it long after it's over. It's like an application re-architected for auto-scaling; it's the same application, but infinitely more powerful.
Today cloud applications are like many of those early talkies, finding their way and using the medium in the best way they can. And just as films evolved from The Jazz Singer, which is essentially a silent film until long after the movie starts, when Al Jolson finally says, "Wait a minute, wait a minute, you ain't heard nothin' yet!" cloud applications are still largely finding their way.
One day, Boris Renski points out, there will be an entire range of OpenStack and private cloud delivery mechanisms we haven't even imagined at this point, just as Fritz Lang couldn't have imagined the sound design of today's blockbusters -- much less the Dolby Digital sound, CGI, web streaming, and other technologies that have changed the face of film in the last century. What new innovations in cloud and cloud delivery will we see as talented architects and engineers and companies figure out new ways for cloud to add agility to their work?
In other words, when it comes to cloud and OpenStack, you ain't heard nothin' yet.