Thursday, April 16, 2009

I Really Don’t Know Clouds at All (video)

As the well-known cloud-architect, Joni Mitchell, said so presciently:
"I’ve looked at clouds from both sides now
From up and down, and still somehow
It’s cloud illusions I recall
I really don’t know clouds at all”
Or, as Larry Ellison put it more succinctly, "What The Hell Is Cloud Computing?"

Like all things Web 2.0, there's an overabundance of fascination with what can be done vs. how fast it can be done or how many things can be done, before the system might fail to scale.

Excerpted YouTube video: In this interview at the CMG 2008 conference, TeamQuest's director of product management, Scott Adams, and I discuss this oft-overlooked aspect of the cloud-computing hype and reflect on what is really needed for better performance and capacity management. My CMG papers on performance management requirements for virtualized services (referred to during the interview) are:
  • "The Virtualization Spectrum from Hyperthreads to GRIDs," CMG 2006 (PDF):
  • "Visualizing Virtualization," CMG MeasureIT 2007 (HTML).
Full length video: If you haven't had enough, the full-length video can be seen at TeamQuest. The downside is, that video is about 20 minutes long and not streamed, so it could take up to a couple of minutes to load. The upside is that some of the most pithy statements about performance requirements for scalability are contained in the second half of this video. Once loaded, you can skip to 8:50, which is where the U2b interview ends.
For better cloud performance and capacity management, we need more whistles and fewer bells.
Update: Lest I may have left the impression here that I am somehow anti-cloud (which is not true), take a look at a my remarks on an early and very impressive cloud-based GIS application of Amazon's EC2, thrown together in a very short time (24-48 hrs?) to search for the late Jim Gray; then missing at sea. Google provided thousands of satellite images, but only Amazon was able to load them quickly into a relational database and front-ended it with their Mechanical Turk interface for literally thousands (perhaps millions) of eyeballs to scan for Gray's yacht. I think this is a remarkable application of cloud technology, which could not have been done in any other way as effectively and efficiently. It also performed well. Moreover, since Amazon runs a real business on such highly distributed systems, it was in a superior position to implement this app in short order. They deserve a lot more kudos than they seemed to get publicly for pulling it off.

1 comment:

William Louth said...

Hi Neil,

Great interview. I really appreciate your effort in pushing for chip/virtualization/cloud vendors to make available much more relevant counters/measures suitable for performance & capacity planning.

I do see a positive side to the current hype surrounding cloud computing in that we are seeing requirements for a more integrated performance & cost management framework. The cost of performance and capacity is much more visible because of the cloud computing billing and it is a much finer granularity than the chargebacks schemes currently being used to some degree in the industry.

I have published a few blog entries on how activity based costing (ABC) can be applied up and down the software execution stack - from a line of code to the execution of a VM or cloud platform.

A Unified Approach to Performance Management and Cost Management for Cloud Computing

Blog: Applying Activity Based Costing (ABC) to Cloud Computing In Practice

Cloud Computing - A Tale of Two Machines

Cloud Modeling with Activity Based Costing

Regards, William