As a result, researchers have been investigating ways to link their cloud deployments with other clouds. Called sky computing, this model takes resources from multiple cloud providers and pools them to create large-scale, distributed infrastructures.
Often, the in-house cloud is linked with resources in public clouds, such as AWS. Sky-computing clouds have emerged at the University of Chicago, Purdue University and the University of Texas at Austin, to name only a few.
Grid5000 in France and FutureGrid in the U.S. are two early sky-computing deployments getting a lot of attention in academic circles. The Ocean Observatories Initiative (OOI), meanwhile, has received $100 million in U.S. stimulus money to deploy sensors and monitoring equipment throughout the worlds oceans. Obviously, handling the data from those sensors will require massive computing infrastructures.
Basically, this is a matter of cloud computing getting back to its roots. Several years ago cloud computing and grid computing were used almost interchangeably. Grids were more typically focused on applications where huge amounts of data where being stored and processed, while clouds were centered on cost reduction and infrastructure flexibility.
Now, sky computing essentially marries the two, with one critical distinction. With a grid, youre still in someone elses domain, Keahey said. You must deal with firewalls and with things being different on different infrastructures. With sky computing, on the other hand, you are able to create a single domain on top of all of these resources.
The next new trend in advertising and marketing, though, is showing promise. According to a new study from King Fish Media, three-quarters of all companies currently have a social-media marketing strategy, and, of those who dont, the vast majority intend to implement one within the next year.
How does the cloud fit into all of this? Simple: the cloud is protecting companies against social-media flops and successes.
In the past, if an expected threshold of users failed to materialize, yet you had already purchased the infrastructure for a large campaign, you were in trouble. Similarly, if a campaign exceeded your projections, you risked alienating potential customers as applications slowed when capacity limits were reached.
Organizations mistakenly believe that big, dedicated enterprise applications are the difficult ones to plan for, said Adrian Ludwig, VP of Marketing for cloud computing provider Joyent. The truth is those are fairly easy to design. The hardware is dedicated to the application, the computing environment is a controlled one, and the organization knows how many end users it will have.
Contrast that with an advertising application where the number of users can range anywhere from a few dozen (if the campaign flops) to tens of thousands, hundreds of thousands or even millions.
Even cloud computing is stressed under that kind of varying demand, Ludwig said.
Joyent tackles this problem by virtualizing the layer between the hypervisor and the application. Just as virtualization separated the hardware assets from the OS and app, Joyents SmartOS frees applications from hypervisors and dedicated OSes. Thus, each application running on one of the ported platforms is just another process that gets managed by the SmartOS.
Using Joyents SmartOS architecture, international advertising firm AKQA has launched and successfully monetized a number of viral marketing campaigns, including a social-shopping campaign for the Gap and a match-planning app created for Visa for the 2010 FIFA World Cup.
Apologies to those who suggested including these cloud innovations: collaboration clouds, cloud mashups, HPC, the appification of software and the emergence of cloud operating systems. Maybe next time. Well, not next time. My next cloud computing story, slated for early to mid-October, will look at how the cloud is changing business and work. If you have ideas for that story, contact me at firstname.lastname@example.org or look for my HARO query in a week or two.