wrote on his blog about why he is leaving the company. It's an interesting look at Google's company culture, but there's also an interesting technical nugget in there. "Google's vaunted scalable software infrastructure is obsolete," Prasanna wrote. He emphasizes that the hardware infrastructure is still state of the art, "But the software stack on top of it is 10 years old, aging and designed for building search engines and crawlers."Yesterday former Google Wave engineer Dhanji R. Prasanna
Prasanna says software like BigTable and MapReduce are "ancient, creaking dinosaurs" compared to open source alternatives like Apache Hadoop.
Prasanna blames the state of Google's software stack on it being designed by "engineers in a vacuum, rather than by developers who have need of tools."
If true, this speaks to the strength of open source - or at least of well maintained open source projects. Open source software can be improved by a wide variety of stake holders, but proprietary software will always be shielded from outside improvements. The open source alternatives have surpassed the proprietary versions that Google kept under lock and key, and Google isn't in a position to take advantage of the improvements made by the open source community without making some major infrastructural changes.
Also, if Prasanna's assessment is correct, it would support RedMonk's Stephen O'Grady's thesis that software infrastructure is no longer a competitive advantage. This is particularly relevant as Google markets its App Engine platform-as-a-service. The Register's Cade Metz recently wrote a long piece on Google App Engine as a means of accessing Google's infrastructure. Although the platform has made improvements in the past year, many developers have been unhappy with its restrictions.
Developers have been willing to accept the proprietary nature of the PaaS and its restrictions to access Google's infrastructure. But what if Google's infrastructure really isn't special? Cloud services powered by open services would then be even more desirable.
We've written before that "open" has won against proprietary, at least in rhetoric if not in practice. Thus far App Engine has bucked that trend. But for how much longer?