Can Microsoft and Its Petascale Financial Services Mining Project Succeed

December 1, 2009

The goslings and I were chattering and quacking in Harrod’s Creek. One of our cousins was killed and eaten for an American holiday. What a way for our beloved friend and colleague to go: deep fried in an oil drum behind the River Creek Inn.

As we were recalling the best moments in Theodore the Turkey’s life, we discussed the likelihood of Microsoft’s petascale content mining project hitting a home run. The ideas, as we addled geese understand it, is that Microsoft wants to process lots of content and generate high value insights for the money crazed MBAs in the world’s leading financial institutions.

The project tackles a number of tough technical problems; for example, getting around the inherent latency in petascale systems, dealing with the traditional input output balkiness of Windows plumbing, and crunching enough data with sufficient accuracy to make the exercise worth the time of the financial client. You may find my earlier post germane.

Other outfits are in this game as well. Some are focused on the hardware / firmware / software side like Exegy. Others provide toolkits like Kapow Technologies. Some Beltway Bandits operate low profile content filtering systems for governmental and commercial clients. And there is the old nemesis, Googzilla, happily chewing through one trillion documents every few days. Finally, some of the financial institutions themselves have pumped support into outfits like Connotate. Even the struggling Time Warner owns some nifty technology in the Relegence unit. So, what’s new?

Three thoughts as I prepare to comment about the push into perceived real time processing at the International Online
Show:

  1. The cost of slashing latency with any type of content is going to be one expensive proposition. Not even some governments have the cash to muscle up a serve with terabytes of RAM. Yep, terabytes.
  2. Figuring out what process left another process gasping for air requires some programmers who can plow through code looking for an errant space, a undefined variable, or a bit of an Assembler hack that push when it should have popped
  3. Latency outside the span of control of the system can render some outputs just plain wrong. Delay is bad; bad outputs are even worse.

If you have not been tracking Microsoft’s big initiatives, you may want to spend some more time grinding through the ACM and other scholarly papers such as “Towards Loosely Coupled Coupled Programming on Petascale Systems.  and poking around on the Microsoft Web site. To find useful stuff, I use the Google Microsoft index. If you aren’t familiar with it, check it out here.

I wonder if this stuff will be part of SharePoint 2011 and available as a Microsoft Fast ESP plug in?

Stephen Arnold, December 1, 2009

Yes, oh, yes. Let me disclose to the National Institute of Science and Technology that I was not paid to write this humorous essay. Consider it a name day present. If I am late, that’s latency. If I am early, that’s predictive output.

Comments

Comments are closed.

  • Archives

  • Recent Posts

  • Meta