Robert Scoble notes the rise of “the serverless Internet company” that can launch and run a webwide business through the window of a browser. He writes of a recent conversation he had with Max Haot, the CEO of Mogulus, a site that lets people produce and broadcast video programs:
At one point Max seemed like he was joking around with me when he told me “we don’t own a single server.” I asked him FOUR more times to make sure I heard him right … He nicely and calmly explained that, yes, every server the company owns is actually running on Amazon’s S3 and EC2 services.
The world has changed. Now ANYONE can build an Internet company and get it up to scale. No more spending nights inside data centers trying to keep servers running.
Yes, Robert, you’re not at Microsoft anymore.
What’s particularly noteworthy about Mogulus is that it shows how layers of utility computing services can be built atop a single shared infrastructure. Mogulus runs its business by drawing on computing and storage services provided by Amazon Web Services, allowing it to avoid any capital investment in computing gear. And then Mogulus offers a set of sophisticated computing services to its own customers, including video editing, storage, and transmission, that until recently would have themselves required big investments in expensive software and hardware. Currently, according to Scoble, some 15,000 people have launched their own channels through Mogulus – and none of them could care less that their work is being stored and processed by Amazon.
Think of how much more efficiently computing assets are used in this model – and, equally important, the way it renders IT essentially invisible to the users. IBM is now making a big push to bring this kind of “cloud computing” to large corporations, through its Blue Cloud program. But the big guys are the latecomers to this revolution. if you want to see the future of computing, look at what the little guys are doing.
Nick, we (AdaptiveBlue) are serverless as well and have been thrilled with the decision to use S3.
You can read a bit about our experiences by clicking the link in my name, but there’s an additional important point in all of this:
While it’s true that most consumers could care less “that their work is being stored and processed by Amazon” there’s an entire group who do care that this is the case.
Luckily, they view it as a positive.
Large companies find piece of mind that the data is being secured and managed by a trusted name like Amazon rather than an unknown start-up who’s working late nights in their basement to ensure that the servers keep running. An unexpected benefit, but an important one.
A little web project that I am involved with is going serverless too.
When I started my career, I used to sit next to a 500 node, 1000 CPU data center, crunching the human genome every couple of weeks. Today, if I was starting a biotech company, I’d really take a hard look at my resources. How many servers would always be up, how often would I have to add excess capacity and do the math. In most cases, especially with EC2 etc adding capacity, where moving large (and I mean laaaarge) volumes of data is not an issue, you’ll find few reasons not to adopt the cloud, especially for certain kinds of tasks.
Of course, for IBM, this is a replacement for a failed grid strategy
What am I missing here? Who holds all the cards in this scenario? Little guys with a few weeks worth of dev work? Or the big guys with all the servers (and lots of devs, too)? This is the whole point about why companies like Microsoft are loving every bit of web 2.0 – because it requires massive amounts of data storage and manipulation – something they can handle (and control) better than almost anyone, Google included.
Kip
I’d say you’re missing a lot. Microsoft, for all it’s expertise, does not hold a candle to Google when it comes to managing distributed datasets that are scaled at a massive level. Microsoft is just about beginning to understand the “cloud” (they hired Dan Reed specifically for that purpose) and even classic HPC where they have never been very good.
I’d say AWS will go down in history as one of the shrewdest diversifications in the tech industry. and yes the servers that Amazon buys do come from the IBMs of the world, so everyone wins
Kip, When I referred to the little guys, I was talking about the users more than the suppliers. I think you’re right that the supply side will tend to consolidate into the hands of big vendors. That’s not necessarily good news for Microsoft, though, as it disrupts a very profitable business model with one whose economics remain unclear – one of the reasons that, as mndoci notes, Microsoft has been slow to adapt. As for mndoci’s claim that “everyone wins,” I’m not so sure about that. Google, after all, builds its own servers and relies largely on free software. Nick
Well, the serverless small internet company has been a reality since the late 90s. Most shopping/blog websites can be run on $90 a month high-end shared hosting.
No one *needs* a server – just an environment they can control and run processes in. Everyone’s decided that the right location for that environment is a safe place near the internet “backbone”. That’s concentrated a lot of servers in a few locations – each of them packed with cheapo barely used-hardware.
What we’re seeing is a consolidation of those servers, one that’s going to be very different from firms (like Zopa where I work) who need to demonstrate physical control of their data centres, to the human genome project who need 1000s of CPUs on-demand with no security constraints. And for most applications, it’s not about getting another 20 CPUs online to cope with a surge, those CPUs have to have the right access to datastorage, different processes are bound by different resources, etc.
I doubt that either MS or Google are best placed to exploit this. Their expertise is in managing their own highly specialised infrastructure. Not in servicing 1000s of demanding time-globbling customers each with their own weird badly configured applications. Firms like Rackspace, Akamai & iPlanet might not be household names, but they have been in that business for quite a while. (Akamai offer some interesting services like [http://www.akamai.com/html/technology/edgecomputing.html]… )
(Given that the EC2 is quite expensive [Amazon have to cope with a lot to make it work] I doubt Mogulus saves any money by using it. It’s probably about time-to-market.)
Amazon AWS seems to be geared towards sites that feature video.
But this is great news for small companies and bad news for large companies like YouTube, because this is eroding yet another competitive advantage on the Internet: a scalable architecture. With AWS, it seems like everyone can create the next YouTube without having to front a lot of money.
I wonder if Amazon plans on coming up with a similar service for databases, where small users can build a database-driven site that can scale.
YouTube already uses 3rd party content distribution networks [CDN] to handle its video streaming. You’re not watching anything on YouTube’s servers – they’re all Limelight.com’s. It’s CDNs that created the economics that made YouTube possible, remember they were a very small fish a few years ago.
S3’s outbound bandwidth is a little expensive and they don’t offer streaming. (Intriguingly, S3 does offer Bittorrent seeding…)