An entirely new IT ecosystem has developed in the last few years around "as a service" or utility offerings. You've got Software as a Service (SaaS), Infrastructure as a Service (IaaS) and Platform as a Service (PaaS), and it can be annoying to try to distinguish one from the next. Not to mention that these terms are still evolving and new ones are joining the mix - for example, where does Storage as a Service fit in? Is it a component of Infrastructure as a Service or is it something that stands on its own?
One point is beyond dispute. These “as a service” entities are fundamentally changing the way IT is done, how applications are developed and deployed, and how information architectures are planned. Public offerings are growing astronomically, private offerings are being pumped out by every vendor you can think of, and the new hybrid class coming into its own.
Each of these entities derives its strengths from the cloud, but I’m going to try to avoid using that word here, and not just because Microsoft has absolutely stripped it of all meaning with those new commercials. When you’re thinking about these utility offerings, I think it’s far more important to focus on the benefits or features of each.
In my previous series, we focused on defining “the cloud” through its features, breaking it down to the attributes the term embodies. Fundamentally, you can’t sell “the cloud.” Relying on it in your marketing (as we have been guilty of in the past) can confuse your customers or, worse yet, get you tossed into the rubbish pile of hype.
Instead, these “as a service” definitions serve as much better transport mechanisms to convey the cloud’s core benefits, or five key attributes:
- Self-Service
- Ease-of-use
- “Infinite” Scalability
- Speed and Agility
- Pay as You Go / On Demand
With these in mind, let’s take a stab at a simple - meaning, less error prone (what do you want, I’m a programmer!) - taxonomy of the as a service ecosystem. At the end, I’ll provide some links for further reading.
Software as a Service
With SaaS, instead of installing and maintaining software on a local disk, applications are hosted, delivered, and accessed over the Internet. Users don’t need to be onsite or on their own computer – they can use the app as long as they have an Internet connection. SaaS offerings include free (sometimes ad-supported) apps like Gmail, Google Reader (RSS) and Google Docs. Google by far is one of the biggest hosts, promoters and believers of SaaS. They even ship a laptop (the CR-48) that only boots into a web browser – you must use SaaS offerings to use the laptop.
SaaS includes a massive variety of pay-as-you-go or subscription-based business and consumer applications. Salesforce.com is another one of the biggest players in the SaaS ecosystem (they started with CRM on demand).
The interesting thing about SaaS is that out of all the concepts presented here, it could be called the oldest. The idea of rich, hosted elsewhere, multi-tenant applications has been around since the early 90s. But the latest generation of offerings is a significant step forward in terms of scale, pricing and ease of use. This boom of SaaS offerings probably owes less of an intellectual debt to the concept of rich web applications from the early 90s than it does to the Web 2.0 trend, when we really saw hosted applications and service take off.
Platform as a Service
With PaaS, you’re getting a foundation or skeleton on which you can build applications (most of which are SaaS applications). When you’re building an application, you normally have to worry about managing hardware, how to deploy it, what framework to use, how to test it, how to scale it, how to bill for it and so on. Most PaaS offerings allow you to skip these choices. At the very least they deal with it for you, or provide a variety of options that tightly integrate with their platform.
For example, Google App Engine - this is Google’s platform offering, and it gives you a choice of two programming languages: Python (our favorite) or Java. With the Python offering, you write your application to Google’s application programming interfaces, relying on their data storage system, load balancing, and account management. You end up paying for the resources your application consumes, but it provides a complete stack on which you can deploy. It automatically scales up and allows you to build and forget it.
The same applies to Heroku, which allows developers to write and deploy Ruby-based apps. Heroku applications normally leverage Ruby on Rails, but they offer a massive selection of add on services that act as one-click add-ons for your site. Once again, they deal with deployment, scalability, hardware and all of the other things you would normally need to when building an application.
Heroku - as a community and company and idea - is so strong that Salesforce.com bought the company for a reported $212 million recently. That’s a very serious movement on Salesforce’s part into the PaaS market, versus the company’s more prevalent SaaS offerings (see Force.com). Even Amazon, which is normally the leading IaaS (Infrastructure as a Service) player, is getting in on the PaaS game with the new Elastic Beanstalk offering (and now CloudFormation).
Infrastructure as a Service
This is where things get weird. IaaS (occasionally called Hardware as a Service) can be very broad. Think back to the post where we discussed the very definition of the cloud. We defined it very tightly (intentionally) to focus on the foundational components you need to build an application or platform. This means things like hardware (servers), pure CPU, message queues (Amazon SQS), database (though you have to wonder if a hosted database could be considered SaaS). Disk/Storage as a service definitely falls into this category. If it’s a component in your datacenter, cluster or computer today, then it could – or is – being sold as a service.
Most people would argue (and it would be hard to dispute) that Amazon is the current leader in IaaS. The company offers servers (EC2), storage (S3 and EBS), database (SimpleDB), message queues (SQS, SNS, etc), and even DNS (Route 53). If it’s a component of a data center today, Amazon seems to be selling it under the IaaS banner (even monkeys – if Mechanical Turk counts).
IaaS is also an increasingly crowded field. GigaOM recently counted Amazon Web Services, Rackspace, Microsoft, GoGrid, Skytap, Linode, Slicehost, OpSource, Terremark, Bluelock, Savvis and AT&T as just a few of the offerings in the IaaS space. Also in here are Joyent, AppNexus and even VMWare. Ultimately, though, we’re going to see consolidation in this space (GigaOM agrees).
Where does this leave us?
The goal of all of these services is the eventual displacement of the modern data center as owned by you, the business owner, developer, or general company. Imagine being able to build a brick and mortar business and never have to buy a single server. Imagine building a Facebook and never needing to pay for electricity, host your own source code repository, or buy a piece of hardware.
So where does Nasuni fit into this picture? We have said that our gateway delivers Storage as a Service, but that’s not really its own standalone category (it folds into IaaS). The Filer itself is Infrastructure as a Service (more on that in a moment) that builds off of Infrastructure as a Service. I know that sounds weird.
The Filer is local infrastructure that adds features and value, building on and leveraging the attributes of the as a service model. Our gateway is instant on, easy to use and dead simple to manage, and it offers unlimited scalability, nearly instant disaster recovery, and agility. Finally, you only pay for what you use. It takes the storage service offered by IaaS providers and adds security, provisioning, migration, file system snapshots, CIFS and NFS access, and much more.
We aggressively test a number of IaaS providers offering a storage component – Amazon S3, Rackspace Cloud Files, Peer 1 hosting for example verifying each one of them to make sure that they hold up to our standards of quality and performance so you don’t have to. You can even bring your own credentials for internally deployed clouds, such as Atmos and OpenStack.
The as a service model of software, platform and infrastructure is only going to grow larger. We will see a continuing stream of providers added to each category. We will see existing players (such as Salesforce and Amazon) extend into new categories. We’re sitting in the middle of an explosion in this space, and it’s exciting to be part of it.
Additional Links:
- Toward a Unified Ontology of Cloud Computing (pdf)
- Unified Ontology of Cloud Computing – John Willis
- The Cloud (Ontology/Taxonomy) Model – Chris Hoff
- The Cloud service stack – Robert Anderson