Over the holidays, I came across this Digg thread on 1&1, in which an unhappy customer griped about his hosting plan's failure to handle a traffic surge.
After his blog got Digged (Dugg?), the customer began getting '500 Internal Server Error'. Shortly thereafter, the message changed to "Forbidden". A "Customer Compliance Operative" (doesn't that sound menacing?) from 1&1 explained that the account was put on hold for using 41% of the server load. This indicates the need for a dedicated server.
The customer complained that he wasn't running anything besides his Wordpress blog. In response, two commenters pointed out that it's not uncommon for Wordpress blogs on shared machines to hog resources. They suggested that the customer consider a VPS.
But this is crazy! The customer has a blog! How can one single blog need a dedicated server - or even a VPS?? Google's Blogger and Microsoft's Windows Live Spaces each host millions of blogs. Can you imagine any one of them consuming 41% of a whole entire server's resources?
That's when a really obvious thought occurred to me: single user apps are far less efficient than multi-tenant SaaS! In addition to Blogger, consider YouTube - and how much more resources it'd need if every single user installed his own video uploading/converting software in separate web space. Or think about Photobucket - and imagine each of its 30 million customers running individual copies of Coppermine or Gallery.
Shared hosting is much more expensive for service providers, and kind of a hassle for most consumers and businesses. I've installed Wordpress and Coppermine before; I have to say I prefer TypePad and Flickr. I haven't checked out Windows Live Spaces, but it must be pretty high on the ease of use scale. Netcraft says it signed up 1.3 million new accounts in August alone.
So there are good reasons why Web 2.0 services are outpacing the web hosting market. User-friendliness is only part of the story, they've got a lower cost structure as well.
As Microsoft Chief Software Architect Ray Ozzie pointed out, consumer-accepted technologies tend to trickle up in the small business market, and eventually impact enterprise IT. This means companies who run their own stand-alone copies of CRM or other back-office software on dedicated servers could very likely begin drifting off to Salesforce.com and other hosted solutions. Because it'd be easier. And more efficient.
In which case, the only remaining audience in the "install your own apps" market will be developers. This means it's time to make up your mind. You could:
Be like Zoho and go after the "manage it for me" market with a well integrated suite of SaaS apps - or
Be like Amazon Web Services and build a one stop developer community.
You can do both, but you won't be very popular if you did neither. Not even if you cut your prices by half.
This cannot happen on VPS because with that architecture each user has resource limits. If that same site was on a VPS then they would overload their own "tenant space" so to say, and would have to go the dedicated server route.
Basic shared servers are good for sites that are not using excessive resources in any way. Obviously this entire idea definitely needs work. It may be cheaper for hosting companies to have what I call the "basic" shared servers, but the stress that comes to a customer who ends up in the excessive resource usage situation is not good at all.
Depending on how "41%" is defined, with a standard WordPress installation + modest number of plugins, you still need to be a very popular site to use 41% of a single server resource. Using an appropriate caching plugin (like WP-Cache2), it should have no issue handling digging or slashdotting, where usually only a single page on the linked site is hammered.
Also, I think clustered shared hosting is going to overcome the traffic surge problem, provided there are large number of nodes in the cluster to offset the load. Other customers will no longer suffer if their "neighbour" is suddenly getting all the attention. The requests will simply be routed onto other server boxes. Overage charge on the CPU time will also revolutionise how shared hosting is done.
I think you are right saying single user apps are far less efficient than multi-tenant SaaS. It is a costly exercise to scale the system up, but it can also have huge overhead to scale the system down. Most SaaS solutions are designed with scalability in mind to have very efficient concurrent access. On the other hand, most web-based FOSS apps are designed to have low overhead in the stock form, so they can easily deployed on shared hosting accounts. Many of them can scale up, but not without appropriate tuning. Automattic's WordPress.com for example -- you probably cannot get that highly scalable well optimised WordPress in a one-click install script...
Then we also have problems with self hosted applications. I don't agree that "install your own apps" will only be for the developers, and SaaS solution is really not for everybody. Take blogging tools for example, you might not agree with the T&C of Google/Microsoft/SixApart, or they don't provide the functionality you want (there are about a zillion FOSS blogging software out there doing different things), and most importantly, many prefer to have their data on their own servers or shared hosting accounts to prevent vendor lock in and other privacy issues.
For example I worked for a company that is closely related to the Australian stock exchange (ASX), and there is no way we will ever be able to out source our MS Exchange servers.
Funny that Microsoft / Ray Ozzie is mentioned. Over the last couple of years we (a SaaS provider for financial software) has been combating against our competitor building their solution on top of Microsoft CRM, and their argument (+ Microsoft's argument) against us has always been that our solution is hosted but their clients can host the software themselves. How wheel has turned...
At least 2/3rds of the Wordpress blogs I've seen aren't customized at all; they're just using the default template. In which case, why can't their owners just blog on a hosted platform? What are they getting out of clogging up your servers with extra overhead? Wouldn't it be cheaper for you and easier for them if you offered them a centrally managed SaaS app?
Scott - the SaaS market is evolving. Soon it might be possible to have your cake and eat it too. I was reading about OmniDrive, a company that offers centralized storage for data generated through Flickr, Zoho and other hosted services. Cleversafe is working towards a much grander scale of the same vision. Such developments will give users more control over privacy issues and prevent vendor lock-in.
My basic thought is, the world is changing - but the web hosting architecture and business model are still the same as they were 10+ years ago. Does it have to be this way? Will there still be a mass market for web space + single user apps in 10 years?