The Meta Cloud - Flying data centers enter fourth dimension

Extra levels of abstraction.

"I just don't see why you would want an extra level of abstraction," says Intel's Jason Waxman. "I don't see where it gets you."

And that is one of the major reasons why "cloud computing" is not, in my mind, an option. It quite simply doesn't scale. Build in-house, save time and money in the long haul.

I don't get it

I can see how you can rapidly switch the demand for CPU resources to cloud B if cloud A goes offline. But where does the data reside? Unless you've replicated it between clouds A and B (or have some sort of RAID thing going on between clouds A, B, C, D, & E), which must be a substantial extra cost - you're stuffed.

I agree with Jake - like any outsourcing deal, it doesn't make economic sense unless you're not big enough to get the economies of scale yourself.

Bad weather ahead?

Somebody once said "There's no problem in Computer Science that can't be solved by adding another layer of abstraction to it" (I think it was meant with a hint of irony). Like drugs, abstraction moves from being useful to an addiction. In IT it's the triumph of the mathematicians over the engineers.

If anybody want a recent example of where this cloudy murk might take us, then it has obvious comparison points with the melt-down in the world's finance systems. There, a set of opaque set of financial products, ever more distanced from the real economy, lead to a risk of systemic failure which nobody understood, or at least cared to think about. Betting your company on a set of complex, poorly understood infrastructure and APIs without really knowing what it going on is a recipe for a catastrophe. There is another financial system metaphor - the way that it bred crime, fraud and misdirection. We've saw Madoff and Standford manage to avoid detection through the opacity of finance systems until it was too later. In the case of the cloud to come, then imagine what the cyber-criminals might get up to.

We all know what clouds bring, and we also know that you can't see through the damn things. Beware the gathering storms. Trust the World's important business IT to this thing and disaster will happen.

Cloud lock-in

Lock-in is the aspect of clouds that make them pretty pointless.

OK in the old days many shops were locked in to an IBM lamebrain complete with EBCDIC-encoded files - but that was an era in which everything was proprietary - one had to pick a vendor and take your chances with the consequences.

The meta-cloud idea might be good if the services of the meta-cloud company are not essential to your operation - even better if there are multiple meta-cloud offerings and migration from one to another is relatively painless (kinda like the early Unix pitch).

Still gives me the willies thinking of entrusting all the organisation's data to a third party - there are so many ways of getting held to ransom.

all your nuts in one basket

I can't help but think those nuts could be squeezed by whoever has possession of them.

The IT revolution has ended...

The CPU is the engine of the IT revolution.

We stick it in a car when we should stick it in a plane.

Currently we pay different chaufers to drive us down different types of road to destinations decided by the chaufers cos we cant be bothered to to learn to drive or map read.

All the chaufers know where the cash machine is tho.

Now the chaufers want us to stay at home and just let us take money out of our accounts electronically while telling us we've just paid for a nice new car.

Me? I took flying lessons years ago and now happily spend my life flying around doing what pigeons do best!

I've taken a few people flying but they keep all their information in big heavy leather bound documents that they have trouble getting on the plane - their chaufers dont seem to want to help.

Admittedly

... I did not get past the first page (acronym level too high), but isn't this just a grid dressed up like a $500 hooker?

A few points about this...

a) Chris Miller's point "I don't get it." (3 comments in.) He sums this point up so perfectly I won't even repeat it.

b) Is this "meta-cloud" software on YOUR end or their end? Otherwise as they say in the article, "But if you opt for RightScale's meta-cloud, aren't you then locked in to RightScale?". You're replacing a single point of failure (the cloud provider) with a single point of failure (RightScale), if it depends on going to a RIghtScale web site. If it can run on your end, I guess you don't have a single point of failure (well, your own computer.. better keep a backup of that RightScale software!)

I'm honestly not a big believer in cloud services, I do see their uses (the compute-intensive rendering and etc. would be OK.) I'm very skeptical of how wise it is to have some huge database "in the cloud" though, getting a copy of this database could be a) difficult and b) costly. Also, depending on the nature of the database, it places confidential company data fully in the hands of a 3rd party, and relies on them to keep it secure. I'm also skeptical of using "the cloud" as some vital backbone of business, I may be traditional but I would think just running regular servers would be better for day-to-day use, with possible cloud use for extra capacity once in a while, if that combination is workable.. that way if the cloud crashes and burns, you are not screwed.

I think it can work for *some* applications.

If like me you are a SME and don't need your own data floor of kit then you have a similar problem in that you are tied to whichever data centre you use to either colo or rent your physical servers from. If that company provides bad service when you have a hardware issue then you can be just as screwed. I had a servers raid array go belly up with 1&1last year. Despite hours and hours on the phone to the dedicated server support team and a 99.99% SLA the server was off line for 11 days! Shortly thereafter another server I had for the purposes of, geographical diversity and redundancy in the event of primary server failure, went off line for 3 days when The Planet had an "explosion" at their Houston data centre. It never rains...

The beauty of AWS is the speed of deployment. I have instances running that have been stable for 9 months now only one has required a reboot. If I do loose an instance I can start a new one and flash the data from a snap shot in minutes. I tried it and it does work brilliantly. It makes me feel a lot better. But I will still have a problem if AWS goes tits up, despite their multiple zones etc. So I still have the physical servers with The Planet. Replicating data between the 2 nodes of servers actually costs me in AWS fees less than The Planet fee, but then my data volumes are comparatively small.

The idea that I could deploy across clouds is hugely appealing to me. But surely the application control panel would have to be non reliant on Right Scale. Maybe it could use Google Gears to run when Right Scale are off line? (I don't know) If it can be done then the killer app would be being able to copy snap shots across clouds that are compatible with each cloud. That would work for me. The only other issue would be who maintains the ip addresses?

Hmmm

I'm in agreement with most on here with regards Cloud Computing - what's the point?

If you're a company such as Salesforce.com or some other SAAS outfit it at first looks good but then when the cloud vapourises(?) your reputation goes with it so better be the master of your own destiny.

SMB - really? Why would they? Too much faffing around surely?

Large enterprise - can't see it. These guys with compute cycle cravings would be better to have a usable/workable spare cycle scavenger on their 000's of desktops and servers. No data problems then as it's all LAN based. Like someone already said, it's a grid dressed up as a $500 hooker.

Render farms and other CPU intensive users - again, the data transfer issue. I'm sure they'd rather run a stack of blade servers or commodity intel workstations and have genuine control.

Hackers - maybe there's a use for it here?

@Mark

"Hackers - maybe there's a use for it here?"

::rolls eyes::

Please, Mark, for the last time ... Read, learn, catch clues, then (and only then) post.