Google's back-end past its sell-by date

Google’s back end is out-of-date, according to one of its former engineers.

Dhanji Prasanna quit the outfit saying that although the search engine was pretty good, the outfit’s distributed back-end was behind the times.

Writing in his bog, Prasanna said that Google’s hardware and datacenters were the best in the world. But the software stack on top of it is 10 years old is ageing and designed for building search engines and crawlers.

Prasanna worked on Guice, a library “at the heart of nearly every single Java server at Google” so he should know what he is talking about.

He said that Google’s Protocol Buffers, BigTable distributed database, and MapReduce distributed number-crunching platform were “ancient, creaking dinosaurs”, compared with outside open source projects that are out there.

Google has admitted that its MapReduce is getting a bit smelly, but Prasanna thinks the rot has got into the newer Google infrastructure projects such as Megastore too,

Prasanna said that developer tools such as Google Web Toolkit and Closure were “sluggish, overengineered Leviathans” compared to MongoDB and jQuery.

Part of the problem, he moaned, was that Google’s new projects are “designed by engineers in a vacuum, rather than by developers who have need of tools.” We can see how working in a vacuum would limit developers, there is not a lot of room in there and it is dusty.

Prasanna said the time it took to rebuild stuff at Google was too slow.

Blowing his own trumpet a bit, he said that in the short time he had been outside Google he had created entire apps in Java in the space of a single workday. He had prototypes off the ground, shown it to people, or deployed them with hardly any barriers. Unlike working at Google.