Has Cloud Computing Killed The Operating System?
T he Operating System originated as a common set of platform management abstractions, to free software developers from the complexity of managing hardware. To many of us that concept has become somewhat distant and when we think OS we usually think in terms of UI or GUI (Graphical User Interface). Most of the current generation of operating systems have become exceedingly bloated primarily because, their creators have been pushing the notion of the OS as a competitive differentiator for many years.
In the early days computers had a single CPU and the obvious thing to do was to build an Operating System that could run many different applications, preferably at once. This created the need for more functionality to manage resources and also manage the applications themselves. Added to this, many OS vendors started to bundle even more widgets, gadgets and applications to their offerings and hey presto – it takes me several minutes to wait for my computer to start-up anytime I want to do anything, this despite the fact that all I want to do is get at Firefox.
My gut-feeling is that Cloud Computing is about to kill the operating system as we know it. The fatal blow has in-fact already been dealt but the demise will not likely be recognised for some time yet.
Desktop Operating Systems
With increasing adoption of Cloud Computing and SaaS it seems to me that, increasingly the tasks I want to complete on a computer are presented to me in a browser. My browser runs on top of my operating system which has a bunch of features and complexity that I would rather forget about. In order to run my operating system with all its bells and whistles the hardware I have is, no doubt vastly over-specked – yet it still takes a number of minutes for me to get going every-time I start-up (Windows Vista in this case is shamefully bloated and slow).
Google recognised this frustration and in July of this year came out and announced its Google Chrome OS. My understanding of Chrome OS is that, in essence it will be a stripped out and hardened Linux Kernel with the ability to run a browser and little or no extra functionality. Google claim they will have users up and running in seconds and promise an end to security vulnerabilities, viruses and malware. This sounds like a good proposition to me – and it will be free, better again.
Meanwhile Microsoft came out and told the world about its research on Gazelle. Gazelle is a research program Microsoft are running under the title “The Multi-Principal OS Construction of the Gazelle Web Browser”. Its interesting that Microsoft’s approach is still underpinned by Windows 7, which I admit I haven’t seen or tested yet, but I have my concerns about its bloat-factor having observed many new Microsoft OSes over many years.
I am not aware of significant research efforts by other companies into offerings that will compete head-on with Goggle’s Chrome OS or Gazelle, however many Linux OSes are very nicely placed to jump into this changing eco-system and capture significant market share. In my view there are three primary reasons that have, thus far stopped Linux from taking a much more significant desktop market share;
Breaking this down a little bit, most Linux OSes are extremely robust and efficient platforms, the GUI in most instances runs as an added application but in my view is pug ugly. So simply by removing the GUI app and running your choice of browser directly on the OS possibly provides a realistic alternative. As mentioned previously, most of what today’s and tomorrows users require is presented to them in a browser, so why bother with the rest of the OS? I guess this is what Google are doing with Chrome OS – but with independent Linux offerings users are not necessarily tied to a specific browser. The problem of installing mainstream applications is quickly disappearing with browser delivered SaaS models, and finally the branding issue? Perhaps there is room here for Firefox, who have a very significant browser market share compared to the resources available to them to jump in and disrupt the market?
So in conclusion, perhaps the Desktop Operating System isnt whats under threat here, maybe we will see the OS going all the way back to its roots again becoming a set of platform management abstractions. Maybe then its the GUI that is at stake? This could explain why Steve Balmer (Microsoft CEO) recently went on the offensive talking trash about the Google Chrome OS and Apples Safari browser, calling them “rounding errors”. In the same breath he also noted that Mozilla’s Firefox was (in the browser market) the most successful so-far. Evidently what is really upsetting Balmer is Google’s recent announcement of its “Internet Explorer Chrome Frame Plug-in”, this according to Balmer is Google replacing Microsoft’s browser rendering engine without telling you, and he calls it an “unanticipated competitive attack factor”.
Microsoft after-all has reason to be concerned, there is a lot at stake here, in their fiscal year 2009 they had revenues of almost $15 Billion from selling their Operating Systems. 80% of which was received from OEM’s (computer manufacturers who bundle windows on their products). If there were to be any significant shift in this pattern it would have an immediate impact on what appears to be Microsoft’s most profitable business line. Additionally the broad market penetration of Windows Desktops eases the way for many of Microsoft’s other products.
Server Operating Systems
Over on the server OS front there is likely to be an even more interesting evolution over time. While many IT Departments and Hosting Companies have for the past couple of years put significant investment of time and resources into virtualising their server infrastructure, little has changed because most of this virtualisation still uses the principle of running a single Server OS on a single virtual machine (basically swapping bare-metal for VM’s).
On the Cloud Computing front Amazon offers an excellent alternative to in-house virtualisation with its EC2 offering and recent analysis by Guy Rosen estimates the number of Amazon EC2 instances launched daily in their us-east-1 region at 50,000 (bear in mind that this is only one of a handful of regions for Amazon EC2).
However many commentators are predicting that Virtual Machines and even Amazons EC2 are only a stop-gap measure on our way to something completely different. The argument is that the Operating System as the owner of a single servers hardware (either bare-metal or virtualised) will lose dominance over time as the the abstractions of computing, storage and networking that enable resource pooling, multi-tenancy, high availability, dynamic workload balancing and the other benefits that arise from a virtualised infrastructure become a reality.
This concept is already taking root with a number of vendors. Noteably Microsoft’s Azure platform decouples the infrastructure completely from the application layer providing flexible virtualised compute, storage and networking capabilities for developers. Currently Azure only supports .NET but as I understand it, support for Ruby-on-Rails, Python and PHP will be added in the near future. Over at Amazon there are 2 services in this category that are also gaining traction, Amazon SimpleDB and Amazon S3, SimpleDB is a virtualised database service and S3 is a storage service.
With all these compute, networking database and storage resources being presented to us as virtual layers and the availability of other major cloud computing offerings such as Google Apps, SalesForce.com, etc., etc., it is becoming increasingly difficult for me to see where the traditional server operating system fits into the IT puzzle of the future.
As always, your comments are welcome.
7 Comments
Trackbacks/Pingbacks
- HTC Touch Diamond – Mobile Phone Review | HTC Touch Windows Mobile Smartphone - [...] Has Cloud Computing Killed The Operating System? « Finbarr McCarthy [...]
Submit a Comment
I thoroughly enjoyed this post.
I think that in general these trends are underway and that on the desktop, you will see the majority of devices move towards nothing more than a hosting platform for a browser. I can see that there will be some desktop applications of such power and complexity (CAD etc.) requiring local device management at a complexity or performance optimizations that would exceed that of a browser-only platform that may still require a more traditional OS. These would clearly be niche applications and would only be applicable to at most 5% of an organization’s desktop devices.
I find your reasoning on the eventual demise of the Traditional Server OS sound but this conclusion does not sit as well with me. Don’t get me wrong, I think it will happen to some extent, but I do not think it will be the same kind of “clean sweep” in the server space as it will be in the desktop space. I understand the attraction of pushing device management to what today we would consider infrastructure devices. A “smart” storage array that presented a secure, sharable programmatic interface to a program running in a virtualized container would eliminate most of the custom I/O drivers for a compute platform. All you would need is a robust network stack and you could access any other services or resources you need. But that is the key – the network.
Performance in this generic, abstracted environment would likely not be as good as the performance offered by a traditional OS. Even with Moore’s Law, the overhead in such an environment would be significant. For all the talk about “bloated” and “cancerous” OS’s, the latency in this kind of loosely-coupled system would be significant. Transaction throughput and latency could be compensated for by increased parallelism in some cases, but for some workloads the additional overhead will be unacceptable. One example is High Performance Computing. HPC is a niche at present and it will likely not be as amenable to these improvements in abstraction if they come at the expense of performance.
There are also certain categories of applications such as real time process control apps in factories, terminals for tellers in your local bank branch or even the system that runs the terminals in your local restaurant. These apps are absolutely intolerant of network outages and so will need to find some means of operating “off the net” if required. That is why I think that the traditional OS will morph into a niche product that runs at the Edge Of the Network (EON) to support branch location operations. The cost of deploying “smart” infrastructure devices at small scale might be too prohibitive to justify. That would still preserve a niche for the traditional OS to run “dumb” devices at very small scales.
To your point, I can envisage designs for future platform services where portions of the central cloud services are replicated locally and engineered to operate autonomously in the event that connectivity to other centralized application services is lost. I’m thinking of something along the lines of SoftGrid application virtualization. Smarter technicians than I can come up with designs far more imaginative.
Again, I found this article enjoyable and thought provoking. I would welcome your thoughts on these observations.
as well with me.
I always enjoy learning what other people think about Amazon Web Services and how they use them. Check out my very own tool CloudBerry Explorer that helps to
manage S3 on Windows . It is a freeware. http://cloudberrylab.com/
Hi Randy
Thanks for your comments; you raise a number of very valid points. In general terms while writing this I was thinking about the majority of personal users, and your typical knowledge worker. There will indeed be several applications and environments in the desktop space that will probably continue to require a more traditional approach for the reasons you mention (network performance & reliability being top of the list).
Speaking about CAD – I read somewhere that Carl Bass (Autodesk CEO) made reference to research Autodesk were conducting to embrace compute power on the cloud for doing 3-D rendering. He explained that the issues AutoCAD users had, was that they didn’t want to wait overnight for 3-D drawing to be rendered – “they want it now” and that elastic cloud compute resources might come to the rescue. However in the same interview he did say that Autodesk was unlikely in to become a SaaS provider.
There are obviously many challenges facing us on network performance and the need for 100% uptime in a model where everything but the browser is running in the cloud. It might take years to get to the stage where we have robust enough networks to consider this mechanism for critical applications – however I believe it will come in time.
I also agree with you that we are unlikely to see a clean sweep across the Server OS space. For one thing, there is the fact that the majority of people currently in charge of provisioning server applications and resources have significant comfort and experience in the existing technologies, however there will probably come a tipping point when both the functionality and price point of services like Amazon AWS, Microsoft Azure and others yet to enter the market will become so compelling that all but the most demanding server apps will be handled by specialists on the cloud (I think the price is compelling right now, but some of the functionality/flexibility is still absent). Once this happens I think the vendors in question will probably employ a whole bunch of specialist OSes on a function by function basis.
Another major influencing factor here will be the “Microsoft Effect”. I think Microsoft has, in some areas been asleep at the wheel and several other vendors have been picking away “softly” at their dominant position in a number of areas. However, Microsoft has proven time and time again to be the “King of the Come-Back”. The challenge for Microsoft will be how to replace $15 Billion of revenue (from their Windows stable) when in all likelihood they will need to give away their OS in the future for free. Perhaps this is why the Azure project is seen by Microsoft execs as being so critical.
I think this is going to be an exciting space to watch over the next few years.
Kind regards
Finbarr
Very enjoyable post. I like your view points. I think the future will (as it is usually is) a blend of many different alternatives. I think the GUI of the OS is bound to change as well as having a more abstract layer to its functions. The notion of more centralized control over “desktop” functions will prevail to justify a “personalized” global view – regardless of which O/S you are on. Anyway I read that the cloud computing market has tripled this year (depends on what you call cloud computing in the first place) so all indications are that it is here to stay and do damage – lets hope its all for the better.
Great article.
As the cost of hardware comes down, and the ability so sync data increases, I think we’ll see the “killing” of the all-purposed computers. OS’s are clumsy when they try to do everything. Devices specially made to do certain tasks are much more efficient. People don’t read books or watch TV at a computer desk for instance.
The new Microsoft Courier device is a great example, and I hope to see more of these single-purposed tools soon.
You sound like you haven’t actually looked at Linux much.
KDE 4.3 is a very slick user experience (for Linux, BSD, UNIX, Mac OS/X and now even Windows XP), and on par on high end machines w/ Win 7 ultimate, and on lower end hardware or lesser versions of Win 7 far better. Memory and CPU use are much less than w/ Win7 or XP, and fancy composting effects which had to be dropped from Vista (and which were excluded from Vista since some Microsoft couldn’t deliver w/o regular BSOD). The eye candy and GUI effects from the composting desktop is really, really cool. I love the effect where I alt-tab and all the open windows shrink and scale so I can see not the icon but the actual open window, just in minature (complete with video playback continuing!).
Mainstream Windows software (except that specifically designed by Microsoft to NOT work) runs fine under WINE, and .NET applications can run using the Mono project. Of course, you can also use Lotus Symphony or Open Office we are feature for feature comparable to any recent version of MS Office. KOffice is another options which I haven’t used much, but what I have (esp. their database program–handles
Mozilla software, Google desktop/earth/talk, Eclipse, Adobe Acrobat/Flash, Java, MySQL GUI tools, and Skype all have Linux versions that work just ducky.
There are thousands of titles I can install with a single click (for the price of, oh, wait, it is free), including a huge amount of kids / educational software, and some serious powerful data analysis, statistics, 3D visualization, video and graphic editing (GIMP is everything Adobe Photoshop is except expensive).
You can get better support via the mailing lists than you will ever get from Microsoft (Although a bit quirky–if you post a request for help, you will get 5 or 6 solutions, followed by some good natured banter about who’s solution is best. There is a huge amount of community good will, and people help others freely).
Security and stability are, well they never have been Microsoft’s strong point, dramatically better. I have had to reboot my desktop PC every few months to install kernel upgrades/patches, and don’t recall the last system crash (it happened to me a few years ago, I think). People have been trying to write viruses for Linux for years, and it is just too hard. Since most hackers do it for the kicks and prestige, creating a virus that would work on Linux would be a major coupe. It has nothing to do with everyone using Windows, and everything to do with the fact that you just cannot get a virus to work in Linux/BSD/other UNIX.
Seriously, give it a few day try, and see for yourself.
Try OpenSUSE (Novell) 11.1 w/ KDE in a VM or live CD/DVD and see for yourself. You can get OpenSUSE to run on just about any hardware, and I only have had one peripheral that didn’t work (a very old “Windows” printer–HP, Epson and Lexmark all have Linux drivers, Cannon isn’t as good about it, but most printers and scanners work). I do, however, stick to HP printers and they do have great Linux drivers.
I might consider using Win7 on my laptop, since it is a pain switching between a single display and an external projector w/ Linux (I have always used ThinkPads so I have never had a problem running Linux on them).
Hi Finbarr, great article.
Nowadays there are basically 2 types of OS, desktop and server.
I suppose you make a good point saying that cloud computing is killing the desktop OSs, probably lightweight OS will be all the rage, effectively making domestic computers much like thin clients.
Also the moment people will understand that a fast reliable office computer already exists, it’s called Linux or FreeBSD with OpenOffice on it, is closer than it seemed only a couple of years ago, so that might be a variable in the desktop OSs scene.
On the server side, given the hype that virtualization is getting these days (most of it justified i have to say), maybe we’ll have two types of server OS: host and guest server OS.
The first one geared towards easier management of virtual machines and the second one more similar to current server OSs, maybe even more specialised in running just one thing but really really well. For example we might have a linux distribution that runs tomcat mad fast, or mysql exceptionally well.
A bit what Kurt was talking about when mentioning single purpose tools, single purpose guest-OSs maybe?
Amazon has just released RDS which allows to instantiate databases on the cloud and complete compatibility with MySQL-connected applications, so even though Amazon might be a stop-gap, they like to think they are here to stay and in fairness this last development is really interesting (http://aws.amazon.com/rds/).
Note to LZ, I love FreeBSD and Linux, but I would never compare GIMP to Photoshop : )