Montreal Tech Watch

In a poll which run here for the past 3 months, Amazon AWS and Google App Engine were choosen by MTW readers as the most promising technologies, well ahead of other new technologies with 34% of votes. I’ve grabbed then the opportunity to delve into virtualized and on-demand web hosting, aka cloud computing.

We sat down with Alistair Croll from Bitcurrent, MC of GigaOm’s cloud-focused Structure08 event and track chair for Interop’s Cloud Computing track, to find out about the cloud.

mtw: Back from last week’s cloudcamp, would you say that cloud computing is hype or it’s here to stay?

Alistair Croll: Cloud is here to stay. The economics are too good to pass up. But there’s a ton of hype surrounding it, because everyone is jumping on the cloud bandwagon. Duncan Hill said it best: Cloud computing is self-service IT outsourcing. That’s a good definition, because it recognizes that what’s new is the ability to self-provision.

I just finished a joint Bitcurrent/GigaOm report on clouds ($250 at http://briefings.gigaom.com), and I was surprised at just how much innovation there is within the field.

Amazon web services and also Google App Engine are the most known providers, with mostly similar features and pricing. But do you see a space somewhere that can by occupied by another startup, by offering for instance different service level agreements, prices, technology, or custom services? Or would are we going towards a market driven by scale, and dominated by Amazon and Google?

It’s still open territory. There are two dimensions of cloud: Operational clouds are about letting someone else run things, and here, Amazon is king (though there are lots of others — Joyent, XCalibre, Gigaspaces, Rackspace.com’s Mosso, for example. On the other hand, Development clouds are about making it easier to build things. The extreme example is Ning, which lets you build a social network.

Generally speaking, development clouds lock you in; operational clouds are “open” and portable, but you have to do a lot of the heavy lifting. Because of that, there are companies like Rightscale and Platespin and enomalism that try to make it easier to manage a cloud.

There’s also a class of companies and open source projects that let you build your own clouds. 3Tera is commonly mentioned here, but there’s also the Eucalyptus project and others.

There are still popular web hosting providers such as dreamhost who offer commoditized web hosting plans. Do they still have a place here? How unique exactly is the cloud computing technology? Would it be possible for web hosters to transform their data centers and offer competing services?

Shared hosting means you’re on the same machine. If someone else kills the machine, you die too. Admittedly, it’s a blurry world — I would say, at a minimum, that a cloud has to have virtualization (to put the customer in a “sandbox”) and self-provisioning. Generally, the virtualization runs atop a grid of commodity hardware so it can expand and shrink as needed.

Many companies (Enki is one) have taken 3Tera’s software and made a cloud. But simply offering a Plesk server and a virtual IP on a load-balancer does not make someone a cloud provider.

How wise is it for a company to host their software to a company like Google or Amazon? Isn’t there a problem if those web giants rolls out a competing product?

Some have claimed that Google Apps is a “farm team” for Google, since you can’t really run your app somewhere else if you want to.  There’s always a tension between the OS and the apps — think about Microsoft introducing disc compression (which competed with Stac) or remote desktop (which competed with Citrix Winframe.) Heck, once upon a time someone competed with MS Paint!

In general, the market will sort out which things belong to virtualization companies like Citrix Xen, VMWare, and Microsoft; which belong to clouds like Amazon, Mosso, and Joyent; and which belong to the apps on top of them.

But consider this: If you build something and run it yourself, you’ll spend a ton of time running it instead of innovating. If you let someone else run it, at least you get to keep developing new features to stay ahead of them. I’d say for many startups, not using the cloud is a huge mistake.

Are cloud computing meant for temporary computing needs? such as handling traffic spikes or temporary data analytics? or are we going towards complete transfer of server infrastructure?

“Bare metal” clouds like Amazon can spin up a server quickly. Enterprises are using them tentatively, for just the kinds of tasks you mention. Internet companies are building their entire businesses on them (think Smugmug.) Some clouds are even built on clouds (like Rails cloud Heroku, which runs on AWS.)

Eventually, cloud computing will be taken for granted. When you come right down to it, you’re not going to build a data center next to a dam, or construce a nuclear power plant. But Google or Amazon might.

There’s seldom any differentiation from running your own hardware. And many services coders normally need to build — like authentication, or storage, or a friend feed, or a mailing system, or a message queue — are already built, and connected to millions of users. So why would you build your own?

To finish, do you have any data on cloud computing adoption? or pointers to best practices for web startups?

The number of developers on Amazon is approaching 400,000. The data from Amazon Web Services already outstrips Amazon.com. This is a fundamental transformation. Companies are building new applications in weeks, and refusing VC money because they don’t have to buy servers in advance.

Some clouds (Joyent is a good example) offer “accelerators” specifically aimed for development of a type of application, such as Facebook. So if you’re a new startup, you’d better have a good reason to not use the cloud. And even if you have one, the cloud should still be factored in for things like testing, bursty capacity, and backups.

For companies wanting to be a cloud, the answer is to specialize. “Vertical” clouds like Salesforce.com’s focus on a specific industry or application. Salesforce is for CRM; Webex is for communications; and so on. I think we’ll see cloud platforms for video (Nirvanix is doing some of this), or for HIPPA compliance, or for travel, and so on. There are big advantages to tailoring your cloud to a specific industry.

Thanks to Alistair Croll for the opportunity!

I would also like to take the opportunity to invite readers to share what their web Infrastructure is. Going around events and blogs, I know this is one of the least talked about subjects, even though it’s a crucial point for a technology company. So, don’t hesitate to leave a comment on how and where you host your software, and your experience with web hosting providers, be it aws or other services. 

Thanks!

Add to: Digg del.icio.us Facebook Reddit StumbleUpon |