At a meeting of high level technology executives Feb. 10, an audience member suggested to a panel that Larry Ellison's description of cloud computing was basically true. There wasn't anything new in it "other than virtualization. The rest is just water vapor," he said knowingly.
The comment was revealing as a measure of how long-lasting some disinformation efforts can be, even -- as in this case -- when they've been abandoned by their author. Haven't you heard? Oracle is a cloud computing company now and everything, even Real Application Clusters (in Exadata), is new again.
The question was remarkable also, considering the parties that it was addressed to. The panel included Doug Cutting, the originator of one of the first useful pieces of cloud software, Hadoop; Chris Kemp, CIO of NASA and a force behind the federal government's pioneering Nebula cloud; Bobby Johnson, director of software engineering at Facebook; and Andy Bechtolsheim, former advanced server designer at Sun Microsystems and now chairman of Arista Networks, which on its Web site describes itself as a cloud networking company.
If Ellison himself had appeared and made such a pronouncement, these panelists could have pinned his ears back. On Thursday night, however, it seemed to me the panelists treated the question with a gentle respect -- but they did disagree.
The event was the CIO Panel, a working group of the CTO Forum, held at the Ritz Carlton in Half Moon Bay, Calif. The CTO Forum is a non-profit group whose membership comes by invitation only. A recent addition, for example, was Mark Carges, CTO of eBay. The CTO Forum sometimes consents to joint projects with the Stanford Graduate School of Business, such as their CTO Institute, established last August, to bring together the latest research results and testing grounds in leading technology companies. Given the "vapor"-charged question and others that evening, the panel's responses highlighted some key features of cloud computing.
Cutting, currently software architect at Cloudera, a firm that produces a commercial implementation of Hadoop, said the cloud "is not just for solving point problems. You have this ability to do things that you hadn't thought of before. ... If you've got thousands of processors, you can do a whole lot with just a few crude tools."
Said Kemp: "I would definitely have difficulty calling the cloud just vapor. ... There's a new architecture emerging in enterprise applications. It's one that assumes not every node is necessarily available," referring to cloud software's ability to tolerate failures in the hardware underneath it.
Cutting amplified his point later in the evening by saying cloud computing moves the issue of system reliability up to a higher level. Rather than seeking reliability in hardware, hardware failures are assumed to happen and the software running on the server cluster can accommodate that.
Bechtolsheim said cloud data management systems use "a whole bunch of nodes running in parallel to find one piece of information. It functions better than [traditional database systems] real-time analytics," in a reference to the use of systems such as Cassandra, CouchDB, MongoDB, HBase, and other NoSQL systems. "You can process data that you could not process before. The scale is a whole magnitude of difference."
Added Johnson: "You can sample things when there's a lot of data involved. The ability to look at data in a few milliseconds can be very powerful." Facebook users tag the people in the pictures they post; in the future, they'll be able to search across Facebook for a given individual appearing with someone else in the same photo, predicted Johnson.
Kemp at another point said the Nebula Cloud gave NASA a reserve in processing power that it's used to look at old data collected in star observations. I hope I heard him correctly; otherwise, I may be guilty of launching another suspect, Carl Sagan-ish comment. Kemp said NASA's data, previously unused and just sitting in storage, upon re-examination revealed "millions of galaxies that we didn't know were there. We've been able to look at the data through a new lens."
So the cloud incorporates commodity hardware. Failures in that hardware occur with regularity among thousands of servers, but do not disrupt the new style of software applications running on top of it. The cloud offers elastic resources that can rapidly expand to meet the demands placed on them, and apply a new data analytics on a scale previously believed impossible to use, or impossible to use and get results that were interactive with end users during a site visit. There are more precise ways to talk about the special characteristics of cloud computing, but for a start, these distinguishing features are not bad.