Recently the SDForum, a Silicon Valley-based technology and business incubator, hosted an all-day Cloud Computing Symposium. If the presenters at the Symposium are to be believed, cloud computing represents an infrastructure revolution, moving infrastructure use from a capital expense to an operational expense and cutting the overall cost by at least 90 percent.
Of course, one must question the validity of these assertions and, crucially, understand the assumptions and implementations of the examples to really grasp whether cloud computing can deliver on the hype. And hype is the right word. I had coffee recently with a "sell-side" financial analyst (i.e., he analyzes stocks with an eye as to whether they are appropriate investments); he told me that, in his estimation, cloud computing had already reached the silly season, with every product under the sun being branded as a cloud offering.
Nevertheless, there was evidence of tremendous payoff from cloud computing at the Symposium.
The Symposium kicked off with an overview presentation by James Staten of Forrester Research. He noted that cloud computing is today being applied very inconsistently, which is to say that startups have embraced cloud computing, particularly Amazon's EC2 and S3 services, while enterprises have not yet begun to adopt cloud computing to any extent. He did note that one place cloud computing will be adopted in enterprises is by end user organizations in an effort to bypass IT organizations seen as expensive and unresponsive.
Staten identified these four types of cloud computing users and the driving reasons for their use of the cloud:
Startups (cheap infrastructure, low investment)
Entertainment (highly scalable, temporary systems; think movie promotional campaigns)
Small businesses (similar motives to startups)
Enterprises (quick and cheap experimentation)
Interestingly, he pooh-poohed one trend forecast as being high-potential for clouds: internal clouds. Essentially, he believes that enterprises cannot create internal clouds that will be cost competitive with external clouds and therefore will ultimately fail. However, he forecast a bright future for virtual private clouds (akin to VPNs) which will carve out sections of a public cloud and provide cloud services overlaid with enterprise-appropriate security, namespace, and isolation.
So what is the big deal about clouds? Why all the excitement?
Essentially, cloud computing offers three key benefits:
First, it provides cheap and easily available infrastructure provided by a third party which specializes in managing computing resources. In some senses, this is akin to what has been called utility computing: its computing and storage on tap.
Second, these resources are managed by the provider to be "always on" and ready to scale. In other words, the provider manages its resource pool so that there is always sufficient headroom so that increased demand can be met. This is in contrast to self-hosted systems, in which scaling up typically runs into issues of budget approval, equipment ordering, and operations overhead.
Third, this easy availability enables quick response to business demands. For startups, as they offer new services or meet sudden demand, the cloud provider makes it easy to respond. For enterprises (as noted earlier), business units can turn to the cloud for IT resources and bypass corporate IT, which may be viewed as overstretched or unresponsive.
What are some examples?
In the realm of startups, several were present at the Symposium and told their story.
One company, Mashery, shared that they were able to build out their IT infrastructure to something like 1000 machines without a single system administrator! Moreover, the CEO indicated that they operated for two years on a total IT investment of less than $100,000.
Another startup, Sharethis (a social media company whose service can be integrated into other websites by merely embedding a button in the site), told the audience that they had been picked up by a very popular website and had to scale from something like 100 to 3500 machines in a single day. Using Amazon's EC2 service, that was achieved easily. The next day, when things had quieted down, they took down many of these instances. Total investment in meeting this transitory spike was less than $200.