When companies excel at some internal function, they sometimes decide to offer that as a service to other companies, as IBM, HP, and others have done with Procurement, Logistics, and Service BPO (Business Process Outsourcing). In Microsoft’s case, they are using their own vast IT data centers and the computer resources within them to offer additional services (notably Azure, their Platform-as-a-Service (PaaS) offering), to their customers. While not everyone is on the same page about what PaaS and other terms mean, most of the more knowledgeable folks in the industry seem to have settled on terminology that has some logic to it:
- Infrastructure-as-a-Service = virtualized infrastructure, such as CPU cycles, disk space, etc. (See #1 below for the meaning of “virtualized”). Examples include Amazon’s EC2 (virtualized compute services) and S3 (virtualized storage services), EMC Atmos, Abacus Data Exchange, CentriLogic, RackSpace Cloud, and many, many others.
- Platform-as-a-Service = Infrastructure plus the virtualized runtime platform/environment. Examples include Windows Azure Platform, Google App Engine, Force.com from Salesforce.com, and Amazon’s RDS and CloudFront services.
- Software-as-a-Service = Infrastructure + Platform + Application. Examples are Salesforce.com and numerous others, a number of which we mentioned in our recent SaaS vs. Licensed Software article.
It is more accurate to call this an “as-a-Service” framework, rather than a complete cloud framework, because it does not encompass a whole variety of other cloud service “flavors” which we have previously enumerated. Within the IaaS/PaaS/SaaS framework, there are some implicit assumptions to the “as-a-Service” designation:
- They are by nature virtualized, in other words multi-tenant and fungible – i.e. you are sharing those resources with others and you are able to expand and contract your use of those resources because of the virtualization. This is an important distinction, which too many providers ignore when they call their offering “SaaS,” while really they are merely hosted.
- The underlying resources are managed by the service provider – i.e. they are responsible for the installation, maintenance, repair, backup, security, etc.
Purpose-built Platforms Prevail
One thing has become apparent – the largest Platform-as-a-Service offerings have grown out of platforms created initially to support the purpose of the organization they served. Google was already running huge data centers with massively scalable databases before they introduced GAE; Saleforce.com was first and foremost a SaaS app, which they then made its underlying platform available to developers; Amazon had already built out a very robust and scalable infrastructure for its own e-commerce business before rolling out the AWS suite. Since so many resources were expended to become absolutely the best at what they do, these firms decided to make additional revenue from those internal capabilities by selling these services externally.
In Microsoft’s case, their Azure offering leverages the massive set of datacenter resources that Microsoft needed anyway for their internal use. At the MIT session, Barry Briggs rattled off some mindboggling figures about Microsoft’s own operations: 4,000 IT employees supporting 80,000 Microsoft employees, 172,000 PCs, 228,000 SharePoint sites, 5 data centers, 7,000 production servers, 188,000 end users, over a million devices, 96 million IMs per month, 1.2 billion hits per day on Microsoft.com, and downloads averaging 300Gb/second. Is your head spinning yet?!?
The On Demand, Dynamically Scalable Nature of “as-a-Service” Resources
Barry also brought up what he called the “seasonality of applications.” For example, certain applications within Microsoft (and many corporations) need lots more CPU power at the end of each quarter, requiring a steep, temporary ramp in resource requirements for that application. When that application runs on a “virtual server” instead of a physical server, it is able to dynamically scale up its use of server resources in the cloud. The same could be true of public website applications that can experience sudden huge spikes in traffic based on events of the day, such as being on the national news.
Barry pointed out that the number of servers required by Microsoft in their datacenters actually had been reduced during the past couple of years from 10,000 down to about 7,000 due to their switching to virtualization of the server environment. This reduction in hardware resource requirements is one of the real, tangible benefits of virtualization/cloud computing.
Three Screens and a Cloud
Barry said that within Microsoft they often talk about “three screens and a cloud,” meaning the cell phone, PC, and TV, all increasingly accessing cloud applications and data. He asked rhetorically “now that a billion people have access to computers, how will the next 3 billion people in emerging economies buy their compute resources?” To make systems ultra-inexpensive, these systems may do it all over the web in the cloud. At the same time, we hear CEOs asking, “How many years before I don’t have a data center?” But corporations need to ask what a large migration into the cloud means for their architecture.
This article is part of our ongoing Cloud Series – our effort to bring clarity to these technologies.
To view other articles from this issue of the brief, click here.