How the Cloud Has Changed (Since Last You Looked)

snydeq writes: InfoWorld's Peter Wayner takes a look at the new services and pricing models that are making cloud computing more powerful, complex, and cheaper than it was a few short years ago. 'We get more, but using it isn't always as simple as it could be. Sure, you still end up on root on some box that's probably running Linux, but getting the right performance out of that machine is more complex,' Wayner writes. "But the real fun comes when you try to figure out how to pay for your planned cloud deployment because there are more options than ever. ... In some cases, the cost engineering can be more complex than the software engineering."

Comments

Hipster software is the real problem.

[Anonymous Coward]

The real problem I've countered is all of the goddamn awful hipster-created software that's out there these days.

We have to wade through mountains of Ruby on Rails and Node.js bullshit. Both of those ecosystems are fucking awful. Even small web apps need tens or hundreds of small, poorly-maintained libraries or modules that some schmuck threw together one weekend, put on GitHub, and then promptly forgot about. But a bunch of other schmucks then chose to build upon this shitty library, so now it's a dependency of all of them. So instead of getting real work done, you'll sit there waiting for rubygems or npm to install all of these fucking awful libraries.

Then there's the NoSQL bullshit. And the Docker bullshit. And the git bullshit. It all piles up!

Dealing with the cloud is the easy part. Dealing with the hipster bullshit is what's hard!

Re:Hipster software is the real problem.

[LordLucless]

> And the git bullshit.

Yeah, we never needed this new-fangled version control bullshit in my day. We just email our code changes around the office. Like men.

Re:Hipster software is the real problem.

[geekpowa]

At least the hipsters successfully buried all the needlessly complex crap the prior (my) generation of engineers inflicted upon the world, the whole horrible XML ediface for one: XML, SOAP, XSLT etc. My hats off to them. Fuck I hated XML, and I dance on it's grave. Good also to see insanely heavy weight app frameworks like J2EE slowly slide away too, great stuff hipsters: it's a terrible legacy you are painstakingly superceding, my apologies for my part in creating it.

Maybe it is in our blood, to channel our energies into at least one needlessly complex endeavour. Their baby is HTML apps (which we foisted upon them BTW so not entirely their fault), and the 'web-scale' cloud.

YMMV but I've found unless you are processing volumes that require rows upon rows of servers, not a couple of slots in a rack, bare metal dedicated gear from a decent 'cloud' provider works just fine. Still simple, and from my personal experience performance and reliability smashes my competitors who use shared cloud based gear. Tried Amazon a while back for a non-profit high volume site I helped migrate. Pricing and complexity of amazon did my head in, nice to hear it has only gotten worse since then. Finally got it running and the thing just crashed into the mountain no matter how many resources and elastic ips and other inscrutable voodoo we threw it at, did some benchmarks and figured DB IO was absolutely abyssmal. Switched to a simpler visualized single host (volumes high enough to justify bare metal but incredibly cost conscious so needed to make virtual work), running years without a hitch.

Re:Hipster software is the real problem.

[sexconker]

I believe the AC is complaining about the general lack of quality, substance, documentation, reliability, and function of many of the new and popular "languages" (which are really frameworks). The examples of Ruby on Rails and Node.js are indeed "fucking awful" from the standpoint of getting worthwhile shit done, especially considering the amount of things you have to seek out and bolt on in order to get them to function as you would expect any other framework to (see the references to gems and npm). The "hipster" label seems to fit. These things are popular not because of their quality or utility, but because of their lack of quality and utility. And despite being popular, the people using them think they're unique for doing so. As soon as the users and developers see that they are no longer unique, they abandon ship and look for the next niche thing to claim as their own.

NoSQL is trash through and through. I have yet to meet a dataset worth looking through, ever, that didn't fit into a relational model of some sort. The anti relational model for data is a joke. If there is any information in data it can be modeled, and if any dataset contains more than one piece of the same type of data then you can develop useful relationships. NoSQL literally throws completeness, correctness, and consistency out the window. "Big data" people love NoSQL and the crazy variants used by Google, Amazon, etc. because they just want to dump data in and sell access to it. They don't care if it's correct, complete, or meaningful. Of course, even Google realizes they need ACID for data that matters, and have basically thrown the non-relational model out the window for their own purposes.

I sure as shit am glad I don't know what Docker is. I've heard references to it before (and I think there was a story on Slashdot about "one weird trick to make Docker take 99% less storage space!!1"), but I simply don't care. As for Git, I fail to see why it's needed when other version control systems already exist and work. My guess is that it gained popularity because of how brain dead simple it is to throw something up on github (which has its own problems, as was discussed on Slashdot recently).

Hipster software IS a fucking problem. Shit like RoR and Node.js fucking suck for most things, but idiots try to shoehorn it and adapt it to do what they need instead of using something better-suited to the task. Ruby (and RoR) is good for certain things, and I presume Node.js may have some sort of valid use case, but certainly not what people usually end up using them for. NoSQL is shit, and Git may be useful but the things people use it for rarely are. The bottom line is that there are people who will scream from the top of the mountain about their new project using Node.js and a non-relational database model that's hosted on github by developers using agile and trending on Twitter, but they won't be able to tell you what their project is or does, or why it's better than the existing solutions.

When the process becomes the project, the project becomes pointless.

Re:Hipster software is the real problem.

[Dunbal]

E-mail. Hah. We used to send over the boxes of punch cards on a trolley. Like men.

Re:Hipster software is the real problem.

[FooAtWFU]

Eh. You're right about things being overhyped but you go too far in the opposite direction. Ruby/Rails is an effective combination for low-volume elevated-complexity latency-insensitive web-based software, the kind you might use internally to your business. Node is a useful tool for quickly writing nonblocking servers (much more useful when they use more than simply HTTP). NoSQL is effective when actually have a ton of data and with the right software (i.e. not Mongo) it can provide guarantees to do everything you want that's also mathematically possible on a data set that large.

But yes, while I like Node in theory, I just wish it wasn't in a freaky language like JavaScript. As for NoSQL... once your problem size is actually legitimately huge then you need to do obnoxious things to make everything work, one way or another, no matter what you do... so it really does pay to avoid it if at all possible (e.g. through clever sharding or the like).

And you're really right about how simple Github is - it almost makes up for the complexity of git itself. Git is a useful and powerful tool that is much nicer than the svn and cvs tools it replaced, and having distributed development available like that is quite effective, but you've actually got to bother to try and learn something about it (otherwise please stick with svn or whatever instead of whining about how a few modestly-cryptic commands and the implications of representing commit history as an immutable DAG are so hard to wrap your head around - this should be undergrad stuff and you've no business passing judgement on entire stacks if you can't grok it).

Re:Hipster software is the real problem.

[Zontar The Mindless]

XSLT kicks ass, and it makes my life easier in more ways than I can count. Don't blame the morons using it for stupid pet tricks for which it was never intended.

Re:Hipster software is the real problem.

[Ace17]

A lot of people seem to equate "distributed version control" to "git". This is incredibly short-sighted (See mercurial, bazaar, bitkeeper, darcs. Raise your hand if you're fluent in any other dvcs than git).

Git is the C++ of version control: it's incredibly powerfull, but needlessly complicated. It's the result of piling unrelated features while trying not to break the workflow of existing users.
The issue is, that after monthes of learning to master this complexity, you become convinced that it's necessary. It's not.

And yes, git can also be seen as version control for hipsters. After all, it's designed around letting people diverge from the accepted path :-)

Re:Hipster software is the real problem.

[thegarbz]

I sure as shit am glad I don't know what Docker is.

Why would you write something like that based on one person's comment that it's "hipster"?

Docker is an open source project that solves a very real problem that linux libraries can often be a disaster to management outside of a package manager by providing a small contained environment in which all dependent libraries can sit.

I've seen someone hose an entire system trying to meet dependencies to compile a single program. Broke apt to the point where nothing would install or uninstall anymore. I myself have bashed my head against the wall trying to get a more current version of software working on an outdated system and eventually gave up because getting the right libraries in place would probably have broken more than it fixed. I eventually installed the program in it's own version of Linux inside a VM on the machine. Just like docker, only 300MB larger still.

Re:Hipster software is the real problem.

[blue9steel]

> And the git bullshit.

Yeah, we never needed this new-fangled version control bullshit in my day. We just email our code changes around the office. Like men.

Real men print them out and send them via inter-office mail. That way they can be read while leaning back in your leather chair, drinking scotch and smoking a pipe.

But

[Ol Olsoc]

Am I supposed to think there is anything actually good about the cloud changes?

More complicated.

A trend toward moving toward "bare metal" physical boxes for the computing.

Moving to Cell phone company like who the hell knows what is what pricing.

All I know is that it seems just as likely a catastrophe in waiting, and given the state of backdooring, ain't happening if I'm a business, because every detail would be exposed.

Re:But

[phantomfive]

The primary advantage of the cloud has always been that it's cheaper (than your own hardware).
Maybe now people are choosing it because they don't have the expertise to do anything else, and so prices are rising to take advantage of their ignorance?

Re:But

[tom229]

It's not cheaper. Stop buying the lie. I've had managed service companies flat out tell me that since they started reselling cloud over internal infrastructure, their revenues have tripled. The only party this crap is good for is the guys selling it. That's why they always seem so excited about it, and anyone on the business side of things hasn't stopped prattling on about the cloud for over 5 years. The one exception would be in very small business; under 20 people.

Re:But

[Zontar The Mindless]

For my next trick, I'll produce something I like to call "other people's servers"...

It's going to take some time still

[Sax Russell 5449D29A]

Dedicated (lease) servers go a long way even if you keep idling most of the available capacity. I think the main difference between cloud and dedicated servers is usability. Cloud is easy to use (albeit sometimes a bit *too* easy) and it's also very easy to go scale in every direction. However, a small team of knowledgeable software engineers could easily go for dedicated servers instead and save a lot of money while also making it equally scalable.

I still tend to prefer dedicated over cloud for everything else except massive backups. Backups can be GPG'd on the fly and that pretty much takes care of all the security concerns we have today regarding cloud usage in this use case. All the execs I've talked to also like the predictability we have in our budgeting. I've had the pleasure to work with some great men and women and we've always had a clear vision about the hardware requirements our solutions need now and in the future, which in turn lets us cast budgeting predictions a long way into the future.

Re:The EC2 price list is pages long!

[sexconker]

Segmented content dynamically loaded via AJAX or similar is the devil.
It's the modern "article split across 10 pages for no reason". Give me a view-all link, and I'll wait for it to load.

For SOME content it's understandable (Twitter's result set for a hash tag, for example, but NOT the result of a user's entire tweet history), but the AJAX loading is still the devil. Give me traditional paginated loading as an option as well as a way to reliably trigger more content to load without making me hold the end key.