In an office, the workers open their thin clients and netbooks to start the day. A network connection is provided at every corner, and there is wireless too, so you can work everywhere you want. All applications run on some big systems, and they just connect to it through their browsers. As soon as they authenticate, they have their own data, and their own work, no matter on which system, or which location they logged in. They can even work at home, or on the road, as long as there’s a connection to the network. Without a network, the whole system is down. Sure, there are some caches, and they can work a bit in advance, but for anything not in their cache, they need a network connection. Picture sounds familiar? Did that just describe cloud computing in it’s essence? Or was it a copy from a computer networking textbook about centralized computing?
There is nothing new in cloud computing. Only the term is, though even that is derived from a decades old concept. From the sixties to the eighties, centralized computing was the way to go. Big mainframes ran every application, and employees used the computer by a dumb terminal. The terminal had no to very minimal processing power, and couldn’t do anything without a connection to a mainframe. When computer parts became more common, likely because they were easier, so cheaper to produce, there was a quick shift to decentralized computing. Processing power at the client was easier and cheaper, and there was no longer a single point of failure. Sure, when the servers went down, workers were disconnected from shared services, but their pc did not stop working. They could do everything they wanted, like starting another task, as long as they did not need one of the disconnected services.
To overcome the issue of decentralized data, technologies like roaming profiles (backed by redundant domain controllers) and VPN connections were invented to allow workers to access shared services everywhere, from their own pc’s, or company provided ‘shared’ pc’s. This is the setup you will find in most companies, and has proven to give the best of both worlds. That is, till Google decided to build ChromeOS: a semi-dumb terminal, useless without a connection to Google’s mainframes. It’s like the sixties all over again!
Apparently, Google thinks that now, a roughly 50 years later, there is a market for a system that will prevent work being done when there is no data connection available, for a system that forces you to share resources with everyone else, and a system that is a single point of failure. Will Google offer a SLA where you get compensation when you can’t do anything because something in their network screwed up? Will they provide compensation when you are waiting for a task to complete, on a server clogged by lots of other users? Will they increase their security to block the added probability their services get compromised? Will they invent a way which blocks scamming and social-engineer-hacking to prevent the leakage of all data?
Does your company trust it’s business-critical data to a company who’s main purpose is crawling and opening up data to the public? Where it’s a feature to have voicemails converted to text, indexed, and searchable for the whole world? Where it is a feature to get advertisements specialized on the contents of your data? Or did you think that was just a coincidence, that an advertisement was closely related to that mail you just received? Any company who values it’s business critical data will choose anything but a search engine company to store it. Of course, in-house is not always the best solution, but there are so much better alternatives, where your data is actually valued, and efforts are taken to keep it away from the public eye.
The only people who could benefit from the ancient model of centralized computing are home users. People with no data to hide, no monetary value to lose when things hit the street. But how many pc’s do these people have? What do they need a centralized cloud for? Sure, it’s handy for some personal documents or some school work, but how are you going to play Call of Duty: Modern Warfare 2? Unless you have an internet connection with at least 1 Gb/s up&down, you will need local processing power to generate the sprites. How are you going to run Photoshop? Can you really keep your patience when you have to wait for each action to complete, because other render jobs are being processed for other people? Can you accept the vendor lock-in, where you can only use the application approved and installed by the cloud vendor?
While the analysts are calling Chrome OS the next move in the cloud war, history has already proven that this move doesn’t work. Well, maybe it does work for Google itself, if they intend to give their employees access to only the few applications available in their own cloud, and remember to do lock away their data from the public eye. For every other company, Chrome OS is a watercooler joke, or a re-enactment of a historic period, if you will.
Filed Under: Two Cents
Released: on Nov 25, 2009 under a Creative Commons Attribution-NoDerivs (CC-BY-ND) license