Two days ago Amazon announced its long awaited move into the PaaS space. Elastic Beanstalk is designed to make the deployment and scaling of Java applications on top of Amazon EC2 effortless. (Don’t be fooled other languages will follow soon.)
Platform as a Service
Now some say this last part, the deployment and scaling of single web applications is what most PaaS providers consider their sweet spot – the single reason how they justify their own margin. Be it Heroku, Engine Yard, phpfog, Scalarium, us or almost any of the impressive list of newcomers have one thing in common. Most sit on top of Amazon Web Services. The single most important reason for that is, there really isn’t much alternatives out there. Yes it’s true, there are other IaaS providers around but most of them are still trying to figure out why an API is a must have feature. This comes as no surprise, because the hosting industry isn’t used to innovate in software. For years they simply used Parallels stuff and competed in pricing. Only this time Parallels missed the bus and is way late to the Cloud party. (On a side note, that “Understanding the Cloud” article is probably the worst piece of info I’ve come across in a very long time.)
Ok, so this wasn’t exactly on topic but it’s important to understand because it is the reason why Amazon’s move into PaaS might turn out to be a big problem for some but not all the other PaaS providers. Effectively Beanstalk is free. Yes you still have to pay for the underlying AWS services you use but there is no extra for Beanstalk itself. By contrast all other PaaS prizes include Amazon’s margin plus the PaaS margin. At this point it looks a little as if Amazon is going to squash the PaaS providers much like it did with the “Cloud Management Platforms” when they launched the AWS console. Anyone remember Rightscale? Sure they support more than just EC2 but as we learned earlier none of those other IaaS providers ever really mattered in the first place.
Surely not. I wrote earlier that Beanstalk might be a problem for some but not all PaaS companies around. Heroku isn’t one of them, and not because they’ve been acquired by Salesforce, but because they are multi tenant. Beanstalk is not, in fact it allows you full access to all underlying AWS services including logging in to EC2 instances. That makes multi tenancy quite hard. At least for now, of course that may be subject to change in the future. Sure differentiation by additional services like an add-on marketplace as suggested here does help a little. But on its own it’s probably not enough. PaaS providers better have another competitive advantage. For add-on providers it’s always better to partner with multiple PaaS providers. No matter if those providers support the same or different languages. Because a variety of add-ons is highly latency sensitive it is best to have those add-ons in the same datacenter as the app servers. This basically means in the same AWS region. That’s why everyone needs Amazon and also Amazon needs to be careful not to ruin that evolving ecosystem because their first and foremost interest must be to have the highest possible occupation of virtual machines possible at all times. They need a wide variety of consumers on top of EC2 for that.
Multi tenancy seems to be the way to go then. Because it allows to scale in much finer grained portions and it also allows providers to offer marketing wise very efficient free tiers. Amazon really has no interest in offering multi-tenant itself, because multi-tenancy means less virtual machines are necessary to host the same amount of applications. Good for Heroku and us, one might think. Not so good for the single-tenant ones. Now I know it’s not black and white and Rightscale also is still around and probably doing just fine. Smaller PaaS providers might differentiate by being close to the customer. But close also often means small, which might be a nice place to be but probably isn’t if you’re venture backed. Good thing is from the way Beanstalk is designed right now it doesn’t seem like Amazon is going for multi tenancy anytime soon and as I said they might not want to in the first place, so there still is time. Bad thing though, multi tenancy is the hard part. So better start now.
Not everything is a web app
True, not everything is, but will be. With all the different devices (Smartphones, Tablets, Desktops, Notebooks, Netbooks, etc.) the web is the least common denominator. It’s definitely going to be exciting to see this evolve in the future.